1 Introducing JobVis:

JobVis is designed to allow job seekers to build interactive visualizations and make useful comparisons between companies and industries based on job characteristics they value the most. Through these features, our app aims to provide deeper insights on an industry and company level so that users can assess whether they are aligned with the company or industry’s requirements and if they fit in with the company/industry culture.

Motivation for this project stemmed from our observations that career portals very rarely provided a holistic view of culture, ratings, skills required (etc), at either an industry level or company level. Instead users would have to manually check each job posting to form an understanding of each characteristic. Additionally, we found that many lacked the nuances of allowing users to make tailored comparisons, and if they did it would be in a tabular format (example: Glassdoors Salary). Another major flaw was that portals did not display visualizations that allow users to keep track of their job applications and success rates. As such, JobVis addresses these pain points through graphical representations, filter functions and an improved UI layout that simplifies and adds value to the job searching experience.

To note, we believe this app is particularly important today since Covid-19 has severely impacted the job market around the world, and has left job seekers questioning which industries and companies are still hiring, what skills they should upgrade to improve chances of landing jobs, and what industry/company best suits their requirements. Moreover, we see that companies are switching to utilizing AI to select candidates, thus it has become imperative for job seekers to determine the keywords that companies are searching for in each job posting and address them in either their resumes or in their job interviews.

2 Source of Infomation:

Job postings shown in our app are scraped from US’ Indeed.com (https://www.indeed.com/jobs?q=data%20scientist&l=all&start=0). The website shows a variety of information from the job position title to the company’s ratings and salary values. Since the information is scraped from online, we were only able to retrieve information for 70 pages (i.e. 1046 data points). Moreover, employee reviews was only extracted for Google, since . We also used “Google Skills” dataset from Kaggle as an example to show users how the application will run when specific companies are chosen for analysis and the different data visualizations.

library(tidyverse)
library(rvest)
library(writexl)

#Obtain Position Title, Company, Location, Date Posted and Link of the Job Posting
df=NULL
for (i in seq(0,700,10)) {
webpage <- read_html(paste0('https://www.indeed.com/jobs?q=data%20scientist&l=all&start=',i))
Sys.sleep(2)
df_partial <- data.frame(
Title=webpage %>% html_nodes('#resultsCol .jobtitle') %>% html_text() %>% str_extract("(\\w+.+)+"),
Company=webpage %>% html_nodes('#resultsCol .company') %>% html_text() %>% str_extract("(\\w+.+)+"),
Location=webpage %>% html_nodes('#resultsCol .accessible-contrast-color-location') %>% html_text() %>% str_extract("(\\w+.+)+"),
Date = webpage %>% html_nodes('#resultsCol .date') %>% html_text() %>% str_extract("(\\w+.+)+"), 
Link= webpage %>% html_nodes('#resultsCol .jobtitle .turnstileLink, #resultsCol a.jobtitle') %>% html_attr('href')) 
df<- rbind(df_partial,df)}
Link <-paste0("https://www.indeed.com",df$Link)
df <- df %>% select(-Link) %>% cbind(Link)

#Obtain the Type, Description and Reviews Link 
for (i in 1:nrow(df)){
  page <- read_html(df$Link[i])
  Sys.sleep(2)
  type <- page %>% html_node('.jobsearch-JobDescriptionSection-sectionItemKey+ div') %>% html_text()
  description <- page %>% html_node('#jobDescriptionText') %>% html_text()
  reviews <- page %>% html_node('.icl-Ratings-count') %>% html_text()
  reviews_link <- page %>% html_node('.icl-Ratings-link') %>% html_attr('href')
  if (is.null(description)){desc <- NA}  
  if (is.null(reviews)){desc <- NA}
  if (is.null(type)){desc <- NA}
  if (is.null(reviews_link)){desc <- NA}
  df$Type[i] <- type
  df$Description[i] <- description
  df$Reviews[i] <- reviews
  df$Reviews_Link[i] <- reviews_link}

#Obtain the Ratings for the Company
for (i in 1:nrow(df)) {
  if (!is.na(df$Reviews_Link[i])) {
  page <- read_html(df$Reviews_Link[i])
  Sys.sleep(2)
  Overall_Review <- page %>% html_node('.cmp-CompactHeaderCompanyRatings-value') %>% html_text()
  Happiness_Review <- page %>% html_node('.cmp-HappinessCompanyRating-happinessScore') %>% html_text()
  Work_Life_Balance <- page %>% html_node('.cmp-TopicFilterList-item:nth-child(1) .cmp-TopicFilter-rating') %>% html_text()
  Pay_Benefits <- page %>% html_node('.cmp-TopicFilterList-item:nth-child(2) .cmp-TopicFilter-rating') %>% html_text()
  Job_Security_and_Advancement <- page %>% html_node('.cmp-TopicFilterList-item:nth-child(3) .cmp-TopicFilter-rating') %>% html_text()
  Management <- page %>% html_node('.cmp-TopicFilterList-item:nth-child(4) .cmp-TopicFilter-rating') %>% html_text()
  Culture <- page %>% html_node('.cmp-TopicFilterList-item:nth-child(5) .cmp-TopicFilter-rating') %>% html_text()
  Woman_Rating <- page %>% html_node('.cmp-FairygodbossWidget-overallRatingScore') %>% html_text()
  
  if (is.null(Overall_Review)){desc <- NA} 
  if (is.null(Happiness_Review)){desc <- NA} 
  if (is.null(Work_Life_Balance)){desc <- NA}  
  if (is.null(Pay_Benefits)){desc <- NA}
  if (is.null(Job_Security_and_Advancement)){desc <- NA}
  if (is.null(Management)){desc <- NA}
  if (is.null(Culture)){desc <- NA}
  if (is.null(Woman_Rating)){desc <- NA}
  
  df$Overall_Review[i] <- Overall_Review
  df$Happiness_Review[i] <- Happiness_Review
  df$Work_Life_Balance[i] <- Work_Life_Balance
  df$Pay_Benefits[i] <- Pay_Benefits
  df$Job_Security_and_Advancement[i] <- Job_Security_and_Advancement
  df$Management[i] <- Management
  df$Culture[i] <- Culture
  df$Woman_Rating[i] <- Woman_Rating
  
  } else {
  df$Overall_Review[i] <- NA
  df$Happiness_Review[i] <- NA
  df$Work_Life_Balance[i] <- NA
  df$Pay_Benefits[i] <- NA
  df$Job_Security_and_Advancement[i] <- NA
  df$Management[i] <- NA
  df$Culture[i] <- NA
  df$Woman_Rating[i] <- NA}}

#Obtain the Company's Industry
companyname <- df$Company %>% gsub(' ','-',.)
industry_complete = NULL
for (i in companyname){
  page <- read_html(paste0('https://www.indeed.com/cmp/',i))
  Sys.sleep(2)
  industry <- data.frame(
    Company = page %>% html_node('.cmp-CompactHeaderCompanyName') %>% html_text(),
    industry = page %>% html_node('.cmp-CompactHeaderCompanyName') %>% html_text())
  industry_complete <- rbind(industry_complete,industry)}
industry_complete

df <- df %>% select(-c(Link,Reviews_Link)) %>% left_join(df,industry_complete,by='Company') %>% select(Title,Company, Industry,Location, Date, Type, Description, Reviews, Overall_Review, Happiness_Review,Work_Life_Balance,Pay_Benefits, Job_Security_and_Advancement,Management,Culture,Woman_Rating)

write_xlsx(df,"/Users/salonipatni/Desktop/Indeed.xlsx") #Save as excel so that everyone has the same data

#Obtain Google Reviews
page <- read_html(paste0('https://www.indeed.com/cmp/Google/reviews'))
Rating <- page %>% html_nodes('.cmp-ReviewRating-text') %>% html_text(trim=TRUE)
Review_Author <-  page %>% html_nodes('meta+ .cmp-ReviewAuthor-link') %>% html_text(trim=TRUE)
Review_Location <- page %>% html_nodes('.cmp-ReviewAuthor-link+ .cmp-ReviewAuthor-link') %>% html_text(trim=TRUE)
Review <- page %>% html_nodes('.cmp-Review-text') %>% html_text(trim=TRUE)
Company <- rep(i,length(Rating))
review <- as.data.frame(cbind(Company,Rating,Review_Author,Review_Location,Review))

write_xlsx(review,"/Users/salonipatni/Desktop/Google_Reviews.xlsx") #Save as excel so that everyone has the same data

closeAllConnections()

2.1 Reading and Cleaning the Data

library(shiny)
library(plotly)
library(xml2)
library(gapminder)
library(readxl)
library(tidyverse)
library(tidytext)
library(scales)
library(stopwords)
library(shinydashboard)
library(DT)
library(shinyWidgets)
library(highcharter)
library(countrycode)
library(dashboardthemes)
library(usmap)
library(openintro)
library(highcharter)
library(treemap)
library(fmsb)
library(networkD3)
library(reshape2)
library(wordcloud)
library(textstem)
library(igraph)
library(ggraph)
library(widyr)
library(shinyswipr)
library(shinysense)

#Read in the data

industry <- read_xlsx('Indeed.xlsx',sheet="Industry")
google <- read_xlsx('Indeed.xlsx',sheet='Google Skills (Kaggle)')
review <- read_xlsx('Indeed.xlsx',sheet='Google Review')
world_long_lat <- read_csv('world_long_lat.csv')

#Cleaning the Dataset 
#Industry Dataset
industry$Description <- industry$Description %>% gsub('\r\n',' ',.)
industry$Date <- industry$Date %>% gsub(' days ago| day ago','',.) %>% gsub('Today|Just posted','0',.)
industry$Location <- industry$Location %>% gsub('[0-9]+.','',.) %>% gsub('\\(.\\)','',.) %>% gsub('United States','Remote',.) %>% gsub('New York State','New York, NY',.) 
industry$Title <- industry$Title %>% gsub('Sr.|Sr','Senior ',.) %>% gsub('Jr.|Jr','Junior ',.) %>% gsub('\\*$',' ',.) %>% gsub("DATA SCIENTIST","Data Scientist",.) %>% gsub("Data Scientist 1","Data Scientist I",.) %>% gsub("Data Scientist 2","Data Scientist II",.)
industry$Industry <- industry$Industry %>% gsub('&','and',.) %>% gsub("Computer and Electronics$","Computers and Electronics",.) %>% gsub("Agriculture$","Agriculture and Extraction",.) %>% gsub('Restaurants and Food Service$','Restaurants and Food Services',.)
industry <- transform(industry, Overall_Review = as.numeric(Overall_Review), Happiness_Review = as.numeric(Happiness_Review), Work_Life_Balance = as.numeric(Work_Life_Balance), Pay_Benefits = as.numeric(Pay_Benefits), Job_Security_and_Advancement = as.numeric(Job_Security_and_Advancement), Management= as.numeric(Management),Culture = as.numeric(Culture), Woman_Rating = as.numeric(Woman_Rating))
# Clean the location column
industry$State_ID <- str_split_fixed(industry$Location, ",", 2)[,2]
industry$State_ID <- trimws(industry$State_ID)
industry$State_ID <- substr(industry$State_ID, start = 1, stop = 2)
industry$State_Name <- abbr2state(industry$State_ID) 
# Add Average salary col
industry$Salary_Avg <- (industry$Salary_Max+industry$Salary_Min)/2

#Google Kaggle Dataset
google$Country <- gsub('.*,\\s*', '', google$Location)
google <- google %>% mutate(Country=case_when(Country=='Dubai - United Arab Emirates'~'United Arab Emirates',Country=='Czechia'~'Czech Republic',Country=='USA'~'United States',TRUE~Country))
job_vacancies <- google%>% group_by(Country) %>%summarise(Num_Vacancy=n())
colnames(world_long_lat)[4] <- "Country"
job_vacancies_complete <- left_join(job_vacancies,world_long_lat, by="Country")
job_vacancies_complete <- job_vacancies_complete[-48,]
job_vacancies_complete$VacanciesCount <- job_vacancies_complete$Num_Vacancy

#Google Review Dataset
review$Review_Location <- review$Review_Location %>% gsub('.*\\d+$','',.) %>% gsub('Work from home|At Home','Remote',.) %>% gsub('2563 S 88th ','',.)
review$Review[26] <- 'I worked there through an agency. The company is very good. Google is one of the best companies in its work environment.' #Convert Spanish reviews with English translation (Can be done using Google Translate API)
review$Review[29] <- 'The team is incredible, the leaders and the daily motivation is excellent.' 
review$Review[150] <- 'It is a place of constant learning and development.'
review$Review[154] <- 'Considered the best place in the world to start working.'

#Salary data
salary_data <- industry %>% drop_na(Salary_Min)

#Stopwords
stopwords <- stopwords::stopwords(language='english',source='stopwords-iso') %>% c(.,"it's","i've","they're","you'll",'u.s.','u.s','job','company','companies','cloud','lot','feel','sap') %>% as.data.frame() %>% rename(.,'word'='.')

3 What makes JobVis unique?

JobVis has three main categories of features that sets us apart from conventional job portals such as Symplicity, Indeed and Glassdoors. These features directly target pain points that we found job seekers find most frustrating. Thus, our unique selling points are as listed below:

3.1 1. Fitting into the Culture

Understanding the company/industry culture in terms of its overall review, happiness review, work-life balance, pay benefits, job security and advancement, etc, is pivotal in determining the user’s fit in the company/industry. Most job portals place very little importance on this or simply provide the statistics without any option to compare.

3.1.1 + Radar Charts with Comparing Capabilities (Industries)

Concerns such as job satisfaction and worklife balance are becoming increasingly important for job seekers today. This function will create radar charts that allow users to see how an industry perform against other industries in terms of several rating metrics. There are seven ratings altogether and they are (1) Overall Review, (2) Work Life Balance, (3) Pay Benefits, (4) Job Security and Advancement/Security , (5) Management, (6) Culture and (7) Woman Ratings.

The Industry vs Average function allows the user to compare the ratings of the various industries. The blue area represents the average ratings across all the industries while the red line represents the average ratings of each of the industry. For instance, we see that Restaurant and Food service performs significantly below the average in terms of Woman Ratings while Education is above the average for all its rating components.

The User vs Industry function, helps the user compare his own requirements against the scores of the different industries. The user score is based on a dummy data but in practice, the scores will be computed based on a survey the user does which will calculate the scores of the user which shows how important each rating is to the user. In the example visualization, the pink area represents the user’s score while the red line represents the score of the industry. The user places more emphasis on Pay, Balance, Stability and Management and he can compare this with the ratings of the various industry to assess his fit with the industry. From this visualization, it seems that Education fits and exceeds most of the user’s criteria.

#Summarize the ratings for each industry
df_radar_ind <- industry %>% filter(Industry != "") %>% group_by(Industry) %>% summarize(Overall = mean(Overall_Review, na.rm = T), 
Balance = mean(Work_Life_Balance, na.rm = T), Pay = mean(Pay_Benefits, na.rm = T), Stability = mean(Job_Security_and_Advancement,na.rm = T),Management = mean(Management, na.rm = T), Culture = mean(Culture, na.rm = T), Woman = mean(Woman_Rating, na.rm = T))
df_radar_ind1 <-df_radar_ind[,-1]
rownames(df_radar_ind1)<-df_radar_ind$Industry

## Fetch minimum and maximum of every column of the dataset
colMax <- function (x) { apply(x, MARGIN=c(2), max, na.rm=T) }
colMin <- function (x) { apply(x, MARGIN=c(2), min, na.rm=T) }
maxmin_ind <- data.frame(max=colMax(df_radar_ind1),min=colMin(df_radar_ind1))
ratings_max <- c(rep(5,7))
ratings_min <- c(rep(0,7))
maxmin_ratings <- as.data.frame(matrix(NA,nrow=7,ncol=2))
maxmin_ratings[,1] <- ratings_min
maxmin_ratings[,2] <- ratings_max
rownames(maxmin_ratings) <-c("Overall" , "Balance" , "Pay", 'Stability', "Management", "Culture", "Woman")
colnames(maxmin_ratings) <-c("min", "max")

## Calculate the average profile
average_ind <- data.frame(rbind(maxmin_ind$max,maxmin_ind$min,t(colMeans(df_radar_ind1, na.rm=T))))
colnames(average_ind) <- colnames(df_radar_ind1)

#Radar function(Industry vs Average)
radar_ind <- function(Industries){
  data <- df_radar_ind1 %>% filter(row.names(df_radar_ind1) %in% Industries)
  par(mar=rep(1.6,4))
  par(mfrow=c(2,2))
  for (i in 1:nrow(data)) {
    toplot <- rbind(
      maxmin_ratings$max,
      maxmin_ratings$min,
      average_ind[3,],
      data[i,])
    plot <- radarchart(
      toplot,
      axistype = 2,
      pfcol = c(rgb(0.2,0.5,0.5,0.4),NA),
      pcol= c(NA,2),
      pty = 32, 
      plty = 1,
      plwd = 2,
      title = row.names(data)[i])
  }
  plot
  par(mfrow = c(1,1))
}
Ind_of_interest <- c("Information Technology", "Energy, Mining and Utilities","Education", "Restaurants and Food Services")

radar_ind(Ind_of_interest) 

#Radar function(Individual vs Industry)
User_Results <- c(4.2, 4.0, 4.0, 3.6, 3.8, 3.4, 3)
radar_ind_user <- function(Industries,User_input){
  data <- df_radar_ind1 %>% filter(row.names(df_radar_ind1) %in% Industries)
  par(mar=rep(1.6,4))
  par(mfrow=c(2,2))
  user_data <- as.data.frame(User_input)
  rownames(user_data) <- c("Overall" , "Balance" , "Pay", 'Stability', "Management", "Culture", "Woman")
  user_data <-t(user_data)
  for (i in 1:nrow(data)) {
                toplot <- rbind(
                    maxmin_ratings$max,
                    maxmin_ratings$min,
                    user_data[1,],
                    data[i,]
                )
                plot <- radarchart(
                    toplot,
                    axistype = 2,
                    pfcol = c(rgb(0.8,0.2,0.5,0.4),NA),
                    pcol= c(NA,2),
                    pty = 32, 
                    plty = 1,
                    plwd = 2,
                    title = row.names(data)[i])
  }
  plot
  par(mfrow = c(1,1))
}
Ind_of_interest <- c("Information Technology", "Healthcare", "Education", "Manufacturing")

radar_ind_user(Ind_of_interest, User_Results)

3.1.2 + Bar Charts to Compare Happiness Ratings by Industries

The function allows users to compare the overall happiness ratings in each industry. The Happiness ratings have been moderated to account for the total happiness ratings indicated by all the job roles across the different industries. The ratings are then broken down into different bands, “Very Low”, “Low”, “Normal”, “High”, and “Very High”. The number of roles in each rating category are also shown in the figure.

# obtaining happiness data 
happiness_complete <-  industry[complete.cases(industry$Happiness_Review),]
happiness_count1 <- happiness_complete %>% select(Industry, Happiness_Review)
happiness_count1$Happiness_Review <- as.numeric(happiness_count1$Happiness_Review)

# moderating happiness data based on the mean and scaling to a range of 0 to 100
happiness_count1$scaled  <- rescale(happiness_count1$Happiness_Review, to=c(0,100))
happiness_count1$happiness_labels <- cut(happiness_count1$scaled, breaks = c(0,20,40,60,80,100),labels = c('Very Low','Low','Normal','High', 'Very High'),include.lowest=TRUE)

happiness_count2 <- happiness_count1 %>% group_by(Industry, happiness_labels) %>% tally() %>% na.omit(cols = "Industry")

# plot figure
happiness <- sample(happiness_count2$Industry, 5)
data.happy <- happiness_count2[happiness_count2$Industry %in% happiness ,]

happiness_fig <- plot_ly() 
happiness_fig <- happiness_fig %>% add_bars(x=data.happy$n, y = data.happy$Industry, color = data.happy$happiness_labels, orientation = 'h' )
happiness_fig <- happiness_fig %>% layout(barmode='stack', xaxis=list( showgrid=T, showlegend = F, zeroline=T,showline=FALSE, ticks='',showticklabels=T))

happiness_fig

3.1.3 + Radar Charts with Comparing Capabilities (Companies)

Similar to the industries section, this function allows users to compare individual company ratings with averages and user input ratings.

Through the Company vs Average function, the user can compare the ratings of the various Companies. The blue area represents the average ratings across all the companies while the red line represents the average ratings of each of the company. For instance, we see that Apple, Bayer and 3M performs well across all the ratings while Tesla is under-performing.

In the User vs Company visualization, the pink area represents the user’s score while the red line represents the score of the Company. The user places more emphasis on Pay, Balance, Stability and Management and he can compare this with the ratings of the various industry to assess his fit with the industry.From this visualization, it seems that only Tesla is not able to meet the user’s requirements.

## Create data for radar plot for ratings by Company
df_radar_comp <- industry %>% filter(Company != "") %>% group_by(Company) %>% summarize(Overall = mean(Overall_Review, na.rm = T), 
Balance = mean(Work_Life_Balance, na.rm = T), Pay = mean(Pay_Benefits, na.rm = T), Stability = mean(Job_Security_and_Advancement,na.rm = T),Management = mean(Management, na.rm = T), Culture = mean(Culture, na.rm = T), Woman = mean(Woman_Rating, na.rm = T))
df_radar_comp1 <-df_radar_comp[,-1]
rownames(df_radar_comp1)<-df_radar_comp$Company

# Fetch minimum and maximum of every column of the dataset
maxmin_comp <- data.frame(max=colMax(df_radar_comp1),min=colMin(df_radar_comp1))

# Calculate the average profile
average_comp <- data.frame(rbind(maxmin_comp$max,maxmin_comp$min,t(colMeans(df_radar_comp1, na.rm=T))))
colnames(average_comp) <- colnames(df_radar_comp1)

#Radar function (Company vs Average)
radar_comp <- function(Companies){
  data <- df_radar_comp1 %>% filter(row.names(df_radar_comp1) %in% Companies)
  par(mar=rep(1.6,4))
  par(mfrow=c(2,2))
  for (i in 1:nrow(data)) {
    toplot <- rbind(
      maxmin_comp$max,
      maxmin_comp$min,
      average_comp[3,],
      data[i,]
    )
    plot <- radarchart(
      toplot,
      axistype = 2,
      pfcol = c(rgb(0.2,0.5,0.5,0.4),NA),
      pcol= c(NA,2),
      pty = 32, 
      plty = 1,
      plwd = 2,
      title = row.names(data)[i]
    )
  }
  plot
  par(mfrow = c(1,1))
}
Companies_of_Interest <- c("3M", "Apple", "Bayer", "Tesla")
radar_comp(Companies_of_Interest)

#Radar functions (User vs Company)
radar_comp_user <- function(Companies,User_input){
  data <- df_radar_comp1 %>% filter(row.names(df_radar_comp1) %in% Companies)
  par(mar=rep(1.6,4))
  par(mfrow=c(2,2))
  user_data <- as.data.frame(User_input)
  rownames(user_data) <- c("Overall" , "Balance" , "Pay", 'Stability', "Management", "Culture", "Woman")
  user_data <-t(user_data)
 for (i in 1:nrow(data)) {
                toplot <- rbind(
                    maxmin_ratings$max,
                    maxmin_ratings$min,
                    user_data[1,],
                    data[i,]
                )
                plot <- radarchart(
                    toplot,
                    axistype = 2,
                    pfcol = c(rgb(0.8,0.2,0.5,0.4),NA),
                    pcol= c(NA,2),
                    pty = 32, 
                    plty = 1,
                    plwd = 2,
                    title = row.names(data)[i]
    )
  }
  plot
  par(mfrow = c(1,1))
}
Companies_of_Interest <- c("3M", "Apple", "Bayer", "Tesla")
radar_comp_user(Companies_of_Interest, User_Results)

3.1.4 + Reviews Analysis

The best way for an outsider to understand a company’s culture is by analyzing the company’s employee reviews. In this section, the user can check the most common positive and negative words filtered by the reviewers rating of the company (1.0-5.0), in the form of a word cloud. Moreover the user receives an overview of the emotional sentiments that the users portray through the reviews. The example data visualizations below are of Google.

#Unnest to individual words
review_unnest <- review %>% select(-Company) %>% unnest_tokens(word,Review) %>% anti_join(.,stopwords, by='word') 

#Creating the Word Cloud
review_unnest %>% filter(Rating=='1.0') %>% inner_join(get_sentiments("bing")) %>% count(word, sentiment, sort = TRUE) %>%acast(word ~ sentiment, value.var = "n", fill = 0) %>% comparison.cloud(colors = c("red", "darkGreen"),max.words = 100)

#Create a NRC sentiment graph
review_unnest %>% inner_join(get_sentiments('nrc') )%>% count(sentiment)%>% mutate(percentage=paste(formatC((n/sum(n))*100,3),'%')) %>% group_by(sentiment) %>% arrange(desc(n)) %>% top_n(10,n) %>% mutate(Type=ifelse(sentiment %in% c('fear','disgust','anger','negative','sadness'),'Negative Emotion','Positive Emotion')) %>% plot_ly(x=~str_to_title(sentiment),y=~n,hovertext=~percentage,color=~Type,colors=c('red','darkGreen')) %>% layout(xaxis=list(title='Sentiment'),yaxis=list(visible=FALSE))
#Positive and Negative words (Location)
ggplotly(review_unnest %>% inner_join(get_sentiments("bing")) %>% anti_join(stopwords) %>% mutate(word=lemmatize_words(word),word=case_when(word=='ambiguity'~'ambiguous',word=='effectively'~'effective',TRUE~word)) %>% filter(Review_Location=='Chicago, IL') %>% count(word, sentiment, sort = TRUE) %>% group_by(sentiment) %>% top_n(10,n) %>% ungroup() %>% mutate(word = reorder(word, n)) %>% ggplot(aes(str_to_title(word), n, fill = sentiment)) +geom_col(show.legend = FALSE) + facet_wrap(~sentiment, scales = "free_y") +labs(y = "Contribution to sentiment",x = NULL) +coord_flip())
#Positive and Negative words (Review_Author)
ggplotly(review_unnest %>% inner_join(get_sentiments("bing")) %>% anti_join(stopwords) %>% mutate(word=lemmatize_words(word),word=case_when(word=='ambiguity'~'ambiguous',word=='effectively'~'effective',TRUE~word)) %>% filter(Review_Author=='Trekker Operator') %>% count(word, sentiment, sort = TRUE) %>% group_by(sentiment) %>% top_n(10,n) %>% ungroup() %>% mutate(word = reorder(word, n)) %>% ggplot(aes(str_to_title(word), n, fill = sentiment)) +geom_col(show.legend = FALSE) + facet_wrap(~sentiment, scales = "free_y") +labs(y = "Contribution to sentiment",x = NULL) +coord_flip())

3.2 2. Discovering Skills Requirements

In order for users to succeed in the application round, the user needs to be well versed in the job skills requirement, which include the Degree Requirement, Technical Skills requirement and the People Skills Requirement. Additionally, this section provides users with insights into common skills within each industry or company that they can use in their resumes to boost their chances of receiving job offers.

3.2.1 + Degree Requirements

Before users apply for a job, they may want to see the degree requirements of the company and whether they fall under the same education standards generally sought for. Under most job applications, users have to manually open each job listing to check the level of education and subject of education required for the job; thus through this visualization the spread can be seen more easily. For Google, we will use the data obtained from Kaggle’s “Google Skills” dataset as it is more accurate than the industry Indeed.com webscraped data.

#Unnest the industry description information 
industry_unnest <- industry %>% select(Industry,Company, Description) %>% unnest_tokens(word,Description) %>% anti_join(.,stopwords,by='word') %>% filter(!str_detect(word,'[0-9]+'))

#Unnest Google Preferred and Minimum Requirement columns
google_unnest <- google %>% select(Category,`Minimum Qualifications`,`Preferred Qualifications`) %>% gather(.,Type,Qualifications,`Minimum Qualifications`:`Preferred Qualifications`) %>% select(-Type) %>% unnest_tokens(word,Qualifications) %>% anti_join(.,stopwords,by='word') %>% filter(!str_detect(word,'[0-9]+'))

#Determine the degrees and subjects
degrees <- c('ba','bs',"bachelor's",'bachelors','phd','masters','ms','mba',"master's","undergraduate","graduate")
subjects <- c('engineering','marketing','finance','Arts','Accounting','Law','economics','science','humanities') %>% tolower()

#Industry frequency of degree or subject requirements 
degreesubject_ind <- function(type,industry){
  
  if(type==degrees) {
    industry_unnest %>% filter(Industry==industry) %>% filter(word %in% type) %>% mutate(word=case_when(word=='bs'~'Bachelors',word=='ba'~'Bachelors',word=="bachelor's"~'Bachelors',word=="bachelors"~"Bachelors", word=='mba'~'MBA',word=='ms'~'Masters',word=="master's"~'Masters',word=="masters"~'Masters',word=='phd'~'PhD',word=='undergraduate'~'Bachelors',word=='graduate'~'Graduates',TRUE~word)) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% plot_ly(x=~str_to_title(word),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Degrees'), yaxis=list(visible=FALSE))
  } else {
    industry_unnest %>% filter(Industry==industry) %>% filter(word %in% type) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% plot_ly(x=~str_to_title(word),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Subject'), yaxis=list(visible=FALSE))
  }
}

degreesubject_ind(degrees,'Healthcare') #Shows degrees most frequent in specific industry -- Healthcare
degreesubject_ind(subjects,'Financial Services') #Shows subjects most frequent in specific industry -- Financial Services
#Company frequency of degree or subject requirements 
degreesubject_comp <- function(type,company){
  if(type==degrees){
    if(company=='Google') {
      google_unnest %>% filter(word %in% type) %>% mutate(word=case_when(word=='bs'~'Bachelors',word=='ba'~'Bachelors',word=="bachelor's"~'Bachelors',word=="bachelors"~"Bachelors", word=='mba'~'MBA',word=='ms'~'Masters',word=="master's"~'Masters',word=="masters"~'Masters',word=='phd'~'PhD',word=='undergraduate'~'Bachelors',word=='graduate'~'Graduates',TRUE~word)) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% plot_ly(x=~case_when(word=='PhD'~'PhD',word=='MBA'~'MBA',TRUE~str_to_title(word)),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Degrees'), yaxis=list(visible=FALSE))
    }else{
      industry_unnest %>% filter(Company==company) %>% filter(word %in% type) %>% mutate(word=case_when(word=='bs'~'Bachelors',word=='ba'~'Bachelors',word=="bachelor's"~'Bachelors',word=="bachelors"~"Bachelors", word=='mba'~'MBA',word=='ms'~'Masters',word=="master's"~'Masters',word=="masters"~'Masters',word=='phd'~'PhD',word=='undergraduate'~'Bachelors',word=='graduate'~'Graduates',TRUE~word)) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% plot_ly(x=~case_when(word=='PhD'~'PhD',word=='MBA'~'MBA',TRUE~str_to_title(word)),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Degrees'), yaxis=list(visible=FALSE))
    }
  } else {
    if(company=='Google') {
      google_unnest%>% filter(word %in% type) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% plot_ly(x=~str_to_title(word),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Subject'), yaxis=list(visible=FALSE))
    } else {
      industry_unnest %>% filter(Company==company) %>% filter(word %in% type) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% plot_ly(x=~str_to_title(word),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Subject'), yaxis=list(visible=FALSE))
    }
  }
}
degreesubject_comp(degrees,'Google') #Shows degrees most frequent in specific industry -- Healthcare
degreesubject_comp(subjects,'Apple') #Shows subjects most frequent in specific industry -- Financial Services

3.2.2 + People Skills Requirements

Most job portals do not provide users with a study of the relatively common people skills that companies are looking for in their candidates. As such, it is ordinary for users to always ask employers ‘what traits and characteristics’ are they searching for in job applicants. To reduce the repetitiveness of the question, it is essential for job portals to showcase these attributes. Thus, through our text mining and visualization, users can see which people skills are most required within the industries and companies. The list of people skills used in the app is not exhaustive, therefore more can be added to obtain higher accuracy. For Google, we will use Kaggle’s “Google Skills” dataset as it provides the minimum and preferred qualifications and hence will allow for a more accurate analysis.

#Define People Skills
softskills <- c('integrity','dependability','effective communication','communication','teamwork','creativity','problem solving','critical','adaptability','organization','willingness to learn','empathy','humble','leadership','lead','resourceful','analytical','motivated','interpersonal','curious','driven','research','work ethic','flexibility','attention-to-detail','confidence','presentation','training','management','independent','dedicated','social','enthusiastic','collaboration','passionate','communicate','decision making','quick','friendly','networking','relationship','team building','analytical thinker','storytelling','critical observation','will learn','innovation','logical thinking','troubleshooting','conflict resolution','deal making','inspiring people','mentoring','resolving issues','managing difficult conversations','giving clear feedback','dealing with politics','hard working','business ethics','meeting deadlines','multitasking','persistence','proper business etiquette','planning','reliability','self directed','experience')

#Industry level Soft Skills
industry_unnest %>% filter(Industry=='Healthcare') %>% filter(word %in% softskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(5) %>% plot_ly(type='bar',x=~str_to_title(word),y=~n, hovertext=~percentage, marker=list(color='purple')) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='People Skills'))
#Company level Soft Skills
industry_unnest %>% filter(Company=='Apple') %>% filter(word %in% softskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(5) %>% plot_ly(type='bar',x=~str_to_title(word),y=~n, hovertext=~percentage, marker=list(color='purple')) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='People Skills'))
#Google Soft Skills
google_unnest %>% filter(word %in% softskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(5) %>% plot_ly(type='bar',x=~str_to_title(word),y=~n, hovertext=~percentage,marker = list(color="purple")) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='People Skills'))

3.2.3 + Technical Skills Requirements

Similarly, most job portals do not provide users with information for technical skills. In this section,

#Define Technical Skills
hardskills <- c("analytic solutions","machine learning","predictive modeling","database systems","clinical decision engines", "algorithms", "NLP/ML", "SQL",  "MongoDB","DynamoDB", "R, ","Python","dplyr","GGPlot", "Pandas","OLS","MLE","Machine Learning",  "Decision Tree","Random Forest","AI" , "Visualization","A/B tests set-up","Reporting",  "data visualizations","numpy", "scipy","scikit-learn", "tensorflow","pytorch" , "keras","genism", "vowpal wabbit","Heap.io","Google Analytics","Big Data","Business Analytics","Oracle","Relational Database Management System (RDMS)","Statistical Programming Language","Regression","Decision Trees","K-Means","Tableau","looker","R Programming" ,"Microsoft Office" , "SPSS","No-SQL", "Cassandra","Hadoop", "Pig","Hive", "HPCC Systems","Javascript" , "Java programming","PowerBI","Linux","TensorFlow", "Keras","Shiny","Artificial Intelligence","NLP", "Tesseract","Jenkins CI/CD", "Azure","logistic regression","k-means clustering","decision forests", "JavaScript","Cloud data", "MATLAB","Excel", "Jupyter","Gurobi","agile", "Git","Github" , "Qlikview","Business Intelligence", "supply chain","D3", "big data",'business sense','C',' API', 'Get Requests', 'Push Requests', 'Update Requests','AWS', 'Sagemaker','Power BI','PowerBI','Cognos','suite', 'Business Objects','Amplitude','Mixpanel','Salesforce', 'Qlik','Microstrategy', 'java','business analytics','reporting','C+','C++','perl','php','hadoop','web scraping','text mining','ruby','Word','NLP','unstructured data') %>% tolower()

#Industry level Hard Skills
industry_unnest %>% filter(Industry=='Financial Services') %>% filter(word %in% hardskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(5) %>% plot_ly(type='bar',x=~case_when(word=='nlp'~'NLP',word=='sql'~'SQL',word=='aws'~'AWS',TRUE~str_to_title(word)),y=~n, hovertext=~percentage,marker = list(color="rgba(255, 0, 0, 0.6)")) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='Technical Skills'))
## Selecting by percentage
#Company level Hard Skills
industry_unnest %>% filter(Company=='3M') %>% filter(word %in% hardskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(5) %>% plot_ly(type='bar',x=~case_when(word=='nlp'~'NLP',word=='sql'~'SQL',word=='aws'~'AWS',TRUE~str_to_title(word)),y=~n, hovertext=~percentage,marker = list(color="rgba(255, 0, 0, 0.6)")) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='Technical Skills'))
## Selecting by percentage
#Google Hard Skills
google_unnest %>% filter(word %in% hardskills)%>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(5) %>% plot_ly(type='bar',x=~case_when(word=='nlp'~'NLP',word=='sql'~'SQL',word=='aws'~'AWS',TRUE~str_to_title(word)),y=~n, hovertext=~percentage,marker = list(color="rgba(255, 0, 0, 0.6)")) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='Technical Skills'))
## Selecting by percentage

3.4 4. Track your Application

While most job portals allow the users to see which companies they had applied to, they do not have functions that allows users to track their job application progress. To allow the job seeker to get a quick overview of their application progress and history, we will include a Sankey Network that represents the user’s progress.

3.4.1 + Sankey Graph

The Sankey Network helps the user track his jobs application progress. The nodes represent the actions of both the user and the company while the size of the link represents the number of actions linking each node (For example: Jobs applied to -> No replies 50 means that 50 of the companies the user applied to did not give a reply to the user). The Sankey Network is plotted based on a dummy dataset for visualization purpose. In reality, the sankey network will track the user’s actions (Jobs applied to, Applicant Rejects, Applicant Accepts) as well as the actions of the companies (No replies, Rejected by Company) which ideally will be recorded on the backend side by the app.

#Create dataset for the sankey network
links <- data.frame(
    source=c("Jobs applied to","Jobs applied to", "Job applied to", "Replies", "Replies", "Initial Interview",
             "Initial Interview","Initial Interview", "Initial Interview", "Offer Received", "Offer Received", "Final Interview",
             "Final Interview","Final Interview", "Offer Received ", "Offer Received "), 
    target=c("Rejected by Company","No replies", "Replies", "Rejected by Company ", "Initial Interview", "Rejected by Company  ",
             "No Followup", "Offer Received", "Final Interview", "Applicant Accepts", "Applicant Rejects", "Rejected by Company   ",
             "No Followup " , "Offer received ", "Applicant Accepts ", "Applicant Rejects "),
    value=c(25 ,50 ,25 ,6 ,19 ,10 ,2 ,2 ,5 ,0 ,2, 2, 1, 2, 1, 1 ))

# From these flows we need to create a node data frame: it lists every entities involved in the flow
nodes <- data.frame(
    name=c(as.character(links$source), 
           as.character(links$target)) %>% unique()
)

# With networkD3, connection must be provided using id, not using real name like in the links dataframe. So we need to reformat it.
links$IDsource <- match(links$source, nodes$name)-1 
links$IDtarget <- match(links$target, nodes$name)-1

# Make the Network
sankey_plot <- sankeyNetwork(Links = links, Nodes = nodes,
                             Source = "IDsource", Target = "IDtarget",
                             Value = "value", NodeID = "name", 
                             sinksRight=FALSE, fontSize = 12)

sankey_plot

3.5 5. Additional Features

3.5.1 + Notifications and Messages

The app will send users notifications when they have not finished an application or if the deadline of the an application is approaching. Furthermore, we have added a messages section on the dashboard header to show the user messages that s/he may have received from their career counselor and other job seekers.

3.5.2 + Summary Dashboard

The dashboard provides users with key statistics and an overview of the data that is used in the app. There are 4 components in the dashboard. 1). Summary statistics (overview of the total number of vacancies, industries, companies & locations) 2). A line graph of the activity rate of recent job postings (to see the hiring rate) 3). Favorite shortcuts for users (which will link to another tab in the app upon click) 4). Filter options and a summary table to display the jobs available in accordance with the filters. The filter options allow users to select multiple companies, industries, locations, salary range and minimum review/rating score.

A Snapshot of the Dashboard

3.6 6. Other Considerations

We acknowledge that there are multiple other features that we can add to our existing app.

3.6.1 + CV Scanner

As users values convenience, providing a scanner for their resumes would naturally improve their whole user experience as they now do not have to manually add in the data which reduces the application time. The application will extract information on the skill sets and experiences listed in the uploaded CV and make comparisons against those required by the companies. We can use shinysense feature (that includes using camera) to add this feature into our app.

3.6.2 + Text Networks

In assessing companies and industries’ attributes, a text network analysis could also be used to describe the relationship in which the jobs have so as to understand and connections between the roles in different industries. This section is aimed to help users identify whether they are interested in the respective roles that they have applied for with this insight.

desc_words <- industry_unnest %>% 
  pairwise_count(word, Company, sort = TRUE, upper = FALSE)

set.seed(1234)
desc_words %>%
  filter(n >= 300) %>%
  graph_from_data_frame() %>%
  ggraph(layout = "fr") +
  geom_edge_link(aes(edge_alpha = n, edge_width = n), edge_colour = "cyan4") +
  geom_node_point(size = 5) +
  geom_node_text(aes(label = name), repel = TRUE, 
                 point.padding = unit(0.2, "lines")) +
  theme_void()

3.6.3 + Job Swipe Option

We tried using shinyswipr and shinysense to introduce a new job swipe feature to our app, but due to its laggy behaviour we decided to ditch the option.

These are some issues we encountered: 1). Using shinyswipr package, we were able to create the “swipe effect” where the swipe box containing a random job profile obtained from the dataset can be swiped. However, this only works for a single swipe box without the update event which would update the box content after one box is swiped. With the code on the update event and history table of previous swipes, the app kept on crashing.

2). We tried using shinysense instead. The app didn’t crash this time. Both the swipe box and the history table of previous swipes can be generated properly. However, the swipe box couldn’t be swiped.

Attempt to use shinyswipr - see the swipe box below

Below is the code that we ran:

#UI Code
u <- shinyswipr_UI("job_swipe",
    p(icon("arrow-left"),HTML("Dislike • Like"),icon("arrow-right"),),
    htmlOutput("joburl"),
    fixedRow(
        column(3),
          column(6,
              column(6,
                  h4("Job Title"), textOutput("job_title"),
                  h4("Location"), textOutput("job_location"),),
              column(6,
                  h4("Company"), textOutput("company_name"),
                  h4("Industry"), textOutput("job_industry"),),),
        column(3)),
          br(), hr(),
                  h4("Top 10"),
                  p("This table updates after you swipe"),
                  column(12, align = "center", tableOutput("table")))

#Server Code 
card_swipe <- callModule(shinyswipr, "my_swiper")
    sample_data <- industry[sample(nrow(industry), 1), ]
    output$job_title <- renderText({ sample_data$Title })
    output$company_name <- renderText({ sample_data$Company })
    output$job_industry <- renderText({ sample_data$Industry })
    output$job_location <- renderText({ sample_data$Location})
    
    # Render the table of swipes
    output$resultsTable <- renderDataTable({ appVals$swipes })
    
    # Set up reactive values object
    appVals <- reactiveValues(
        swipes = data.frame(
            job_title = character(), company_name = character(), job_industry=character(), job_location=character(),swipe = character()
        )
    )
    
    observeEvent(card_swipe(), {
        
        # Record the last swipe result
        latest_result <- data.frame(
            job_title = appVals$sample_data$Title,
            company_name = appVals$sample_data$Company,
            job_industry = appVals$sample_data$Industry,
            job_location = appVals$sample_data$Location,
            swipe = card_swipe())
        
        # Add to table of all swipe results
        appVals$swipes <- rbind(latest_result, appVals$swipes)
        
        # Send results to the output
        output$resultsTable <- renderTable({ appVals$swipes })
        
        # Update the villager
        appVals$sample_data <- sample_n(industry, 1)
        
        # Send update to ui
        output$job_title <- renderText({ appVals$sample_data$Title })
        output$company_name <- renderText({ appVals$sample_data$Company })
        output$job_industry <- renderText({ appVals$sample_data$Industry})
        output$job_location <- renderText({ appVals$sample_data$Location})
        
        the_data <- eventReactive(
            { card_swipe() }
        )
        
        output$table <- renderTable(the_data())
        
    })

4 Code for our App

library(shiny)
library(shiny)
library(plotly)
library(xml2)
library(gapminder)
library(readxl)
library(tidyverse)
library(tidytext)
library(scales)
library(stopwords)
library(shinydashboard)
library(DT)
library(shinyWidgets)
library(highcharter)
library(countrycode)
library(dashboardthemes)
library(usmap)
library(openintro)
library(highcharter)
library(treemap)
library(fmsb)
library(networkD3)
library(reshape2)
library(wordcloud)
library(textstem)
library(igraph)
library(ggraph)
library(widyr)
library(shinyswipr)
library(shinysense)
library(shinydashboardPlus)
library(magrittr)
library(shinyhelper)
#devtools::install_github("nstrayer/shinyswipr")
#devtools::install_github("nstrayer/shinysense")

#------------------------------------------------------------------------------------------------------------------------------------

#Read in the data
industry <- read_xlsx('Indeed.xlsx',sheet="Industry")
google <- read_xlsx('Indeed.xlsx',sheet='Google Skills (Kaggle)')
review <- read_xlsx('Indeed.xlsx',sheet='Google Review')
world_long_lat <- read_csv('world_long_lat.csv')

#Cleaning the data 
#Industry Dataset
industry$Description <- industry$Description %>% gsub('\r\n',' ',.)
industry$Date <- industry$Date %>% gsub(' days ago| day ago','',.) %>% gsub('Today|Just posted','0',.)
industry$Location <- industry$Location %>% gsub('[0-9]+.','',.) %>% gsub('\\(.\\)','',.) %>% gsub('United States','Remote',.) %>% gsub('New York State','New York, NY',.) 
industry$Title <- industry$Title %>% gsub('Sr.|Sr','Senior ',.) %>% gsub('Jr.|Jr','Junior ',.) %>% gsub('\\*$',' ',.) %>% gsub("DATA SCIENTIST","Data Scientist",.) %>% gsub("Data Scientist 1","Data Scientist I",.) %>% gsub("Data Scientist 2","Data Scientist II",.)
industry$Industry <- industry$Industry %>% gsub('&','and',.) %>% gsub("Computer and Electronics$","Computers and Electronics",.) %>% gsub("Agriculture$","Agriculture and Extraction",.) %>% gsub('Restaurants and Food Service$','Restaurants and Food Services',.)
industry <- transform(industry, Overall_Review = as.numeric(Overall_Review), Happiness_Review = as.numeric(Happiness_Review), Work_Life_Balance = as.numeric(Work_Life_Balance), Pay_Benefits = as.numeric(Pay_Benefits), Job_Security_and_Advancement = as.numeric(Job_Security_and_Advancement), Management= as.numeric(Management),Culture = as.numeric(Culture), Woman_Rating = as.numeric(Woman_Rating))
# Clean the location col
industry$State_ID <- str_split_fixed(industry$Location, ",", 2)[,2]
industry$State_ID <- trimws(industry$State_ID)
industry$State_ID <- substr(industry$State_ID, start = 1, stop = 2)
industry$State_Name <- abbr2state(industry$State_ID) 
# Add Average salary col
industry$Salary_Avg <- (industry$Salary_Max+industry$Salary_Min)/2

industries <- sort(unique(industry$Industry))

industries <- sort(unique(industry$Industry))

#Google Dataset
google$Country <- gsub('.*,\\s*', '', google$Location)
google <- google %>% mutate(Country=case_when(Country=='Dubai - United Arab Emirates'~'United Arab Emirates',Country=='Czechia'~'Czech Republic',Country=='USA'~'United States',TRUE~Country))
job_vacancies <- google%>% group_by(Country) %>%summarise(Num_Vacancy=n())
colnames(world_long_lat)[4] <- "Country"
job_vacancies_complete <- left_join(job_vacancies,world_long_lat, by="Country")
job_vacancies_complete <- job_vacancies_complete[-48,]
job_vacancies_complete$VacanciesCount <- job_vacancies_complete$Num_Vacancy

#Google Review Dataset
review$Review_Location <- review$Review_Location %>% gsub('.*\\d+$','',.) %>% gsub('Work from home|At Home','Remote',.) %>% gsub('2563 S 88th ','',.)
review$Review[26] <- 'I worked there through an agency. The company is very good. Google is one of the best companies in its work environment.'
review$Review[29] <- 'The team is incredible, the leaders and the daily motivation is excellent.'
review$Review[150] <- 'It is a place of constant learning and development.'
review$Review[154] <- 'Considered the best place in the world to start working.'

# salary data
salary_data <- industry %>% drop_na(Salary_Min)

# Ratings data
happiness_complete <-  industry[complete.cases(industry$Happiness_Review),]
happiness_count1 <- happiness_complete %>% select(Industry, Happiness_Review)
happiness_count1$scaled  <- rescale(happiness_count1$Happiness_Review, to=c(0,100))
happiness_count1$happiness_labels <- cut(happiness_count1$scaled, breaks = c(0,20,40,60,80,100),labels = c('Very Low','Low','Normal','High', 'Very High'),include.lowest=TRUE)
happiness_count2 <- happiness_count1 %>% group_by(Industry, happiness_labels) %>% tally() %>% na.omit(cols = "Industry")

#Stopwords
stopwords <- stopwords::stopwords(language='english',source='stopwords-iso') %>% c(.,"it's","i've","they're","you'll",'u.s.','u.s','job','company','companies','cloud','lot','feel','sap') %>% as.data.frame() %>% rename(.,'word'='.') 

# Create Industry Lists
Industry_list <- list(unique(industry$Industry))[[1]]
Industry_list <- Industry_list[!is.na(Industry_list)]
Industry_list <- sort(Industry_list)
Industry_list_with_All <-append(Industry_list,"All", after=0)
Industry_list_with_NA <- append(Industry_list,"NA", after=0)
Industry_list_with_All
Industry_list_with_NA
# Create Company Lists
Company_list <- list(unique(industry$Company))[[1]]
Company_list <- Company_list[!is.na(Company_list)]
Company_list <- sort(Company_list)
Company_list_with_NA <- append(Company_list,"NA", after=0)


#------------------------------------------------------------------------------------------------------------------------------------

# Define UI for application that draws a histogram
ui <- 
    dashboardPage(
    dashboardHeader(title='JobVis'),
    dashboardSidebar(sidebarMenu(
        
        menuItem("Dashboard", tabName = "Dashboard", icon = icon("dashboard")),
        
        menuItem("User Profile", tabName = "User", icon = icon("address-card")),
        
        menuItem("Jobs", tabName = "Jobs", icon = icon("briefcase"),
                 menuSubItem('Jobs by Industry', tabName = "Jobs_by_Industry", icon = icon('industry')),
                 menuSubItem('Jobs by Company', tabName = "Jobs_by_Company", icon = icon('building'))),
        
        menuItem("Salary", tabName = "Salary", icon = icon("briefcase"),
                 menuSubItem('Salary by Industry', tabName = "Salary_by_Industry", icon = icon('industry')),
                 menuSubItem('Salary by Company', tabName = "Salary_by_Company", icon = icon('building'))),
        
        menuItem("Ratings", tabName = "Ratings", icon = icon("briefcase"),
                 menuSubItem('Ratings by Industry', tabName = "Ratings_by_Industry", icon = icon('industry')),
                 menuSubItem('Ratings by Company', tabName = "Ratings_by_Company", icon = icon('building'))),
        
        menuItem("Skills", tabName = "Skills", icon = icon("code-branch")),
        
        menuItem("Reviews", tabName = "Reviews", icon = icon("comments")),
        menuItem("More information", tabName = "More_info", icon = icon("info-circle"))
        #menuItem("Other Plans", tabName = "Other Plans", icon = icon("address-card")),
        #menuItem("Other Plans", tabName = "Other Plans", icon = icon("address-card")),
           )),
    dashboardBody(shinyDashboardThemes(
        theme = "flat_red"),
        chooseSliderSkin("Flat", color = "#112446"),
        tags$script(HTML("
        var openTab = function(tabName){
          $('a', $('.sidebar')).each(function() {
            if(this.getAttribute('data-value') == tabName) {
              this.click()
            };
          });
        }
        ")),
        tabItems(
            #Dashboard Content
            tabItem(tabName = "Dashboard",
                    tags$strong(tags$h1("JobVis: For a Smarter Job Search",style = "color:black;")),
                    tags$h4("This dashboard will give you an overview of job vacancies across a wide range of industries and companies around the world."),
                    valueBoxOutput("valuebox_vacancies",width=3),
                    valueBoxOutput("valuebox_companies",width=3),
                    valueBoxOutput("valuebox_countries",width=3),
                    valueBoxOutput("valuebox_industries",width=3),
                    
                    fluidRow(column(width=8,
                                    plotlyOutput('job_posting'),height="400px"),
                             box(title="My Favorite Shortcuts",status='primary',width=4,color='red'),
                             infoBoxOutput("out_salary"),
                             infoBoxOutput("out_requirements"),
                             infoBoxOutput("out_jobfilter")),
                    fluidRow(gradientBox(title="Filter Options",icon = "fa fa-th",
                                         boxToolSize = "sm", width=12,solidHeader=T,status='info',
                                         fluidRow(
                                             column(4,selectInput('filter_location','Location: ',choices=c("All Locations",unique(industry$State_Name,na.rm=TRUE)),selected ="All Locations",multiple=T)),
                                             column(4,selectInput('filter_company','Company: ',choices=c("All Companies",unique(industry$Company, na.rm=TRUE)),selected="All Companies",multiple=T)),
                                             column(4,selectInput('filter_industry','Industry: ',choices=c("All Industries",unique(industry$Industry,na.rm=TRUE)),selected="All Industries",multiple=T))),
                                         fluidRow(
                                             column(6,sliderInput('filter_salary','Salary Range: ',min=0,max=ceiling(max(industry$Salary_Max, na.rm = TRUE)),value=c(700,8000))),
                                             column(6,sliderInput('filter_rating','Overall Ratings more than: ',min=0,max=5,value=1)),
                                             tableOutput("job_summary_table"))))),
            
        #User Profile Content
        tabItem(tabName = "User",
                h2("User Profile"),
                tags$h4("View your job application progress."),
                widgetUserBox(width=12,
                    title = "Jane Doe",
                    subtitle = "Business Analytics Student at NUS Business School",
                    type = 2,
                    src = "https://adminlte.io/themes/AdminLTE/dist/img/user7-128x128.jpg",
                    color = "yellow",
                    "About Me",
                    footer = "I am in my sophomore year of university and I aspire to be a business analyst. I did a summer stint as a data scientist intern at Google and I am currently looking out for job opportunities within the tech industry."
                ),
                fluidRow(
                    box(width = 12,
                        title = "User Progress",
                        color = "green", ribbon = TRUE, title_side = "top right",
                        column(width = 5),
                        sankeyNetworkOutput('UserProgress') %>%
                            helper(type = "inline",
                                   title = "Sankey Network",
                                   content = "Each node represents the actions of either the user of the company and the number on the link represents the
                                   value of the interaction",
                                   size = "s"),
                    ))
        ),
       
        #Jobs Content
        
        #Jobs by Industry 
        tabItem(tabName = "Jobs_by_Industry",
                h2("Jobs by Industry"),
                tags$h4("Compare Job vacancies in the US by Industry"),
                fluidRow(
                box(title='Industry of Interest',width=12,status='warning',
                    selectInput('usmap_ind','Choose an industry:',Industry_list_with_All)),
                box(title='Map of Job Vacancies', width=12,status='warning',solidHeader = TRUE, 
                    uiOutput('heatmapUI')%>%
                        helper(type = "inline",
                               title = "HeatMap of the Job vacancies by state",
                               content = "The color density represents the number of job vacancies across the different states by Industry",
                               size = "s"))),
                 
        ),
        
        # JObs by company
        tabItem(tabName = "Jobs_by_Company",
                h2("Jobs by Company"),
                tags$h4("Compare job vacancies across different countries in the world for a company."),
                fluidRow(
                box(width=12,status='danger',
                    selectInput('pickCompany','Choose a company: ',choices=c("Google"),selected="Google")),
                
                box(title='Map of Job Vacancies',width=6,status='warning',solidHeader=TRUE,
                    selectInput('country','Select Country: ', choices=c("All Countries",unique(job_vacancies_complete$Country))),
                    sliderInput('vacancies','Vacancies more than: ',min=1,max=88,value=1),
                    hr(style = 'border-color:black;'),
                    highchartOutput('plotChoroplethMap',height='400px')%>%
                        helper(type = "inline",
                               title = "Choropleth Map of Job Vacancies around the World",
                               content = "The color density represents the nunber of vacancies across the different countries in the world for a company. Hover over the map to see more details.",
                               size = "s")),
                box(title='Filtering Jobs by Location or Category',width=6,status='warning',solidHeader=TRUE,
                    selectInput('continent', 'Choose a continent: ', 
                                choices = c("Asia","Africa",
                                            "Europe","Oceania")),
                    selectInput('filter_type','Choose a filter type: ',choices=c("Country","Category")),
                    plotlyOutput('plotSunburst',height='450px')%>% helper(type = "inline",
                                                                          title = "Sunburst Chart of Job Vacancies by Country or Category",
                                                                          content = "Hover over each sunburst chart to view more details on each vacancy. Click over each slice of the chart to expand and enlarge the view of the vacancies for that particular country/category. ",
                                                                          size = "s")))),
        
        
    
        
        #Salary Content
        ## Salary by Industry
        tabItem(tabName = "Salary_by_Industry",
                h2("Salary and Vacancies by Industry"),
                tags$h4("Compare the average salaries and Job vacancies of different companies based on the industry."),
                fluidRow(
                box(title='Industry of Interest',width=12,status='warning',
                    selectInput('salary_tree_ind','Industry:',Industry_list_with_All)),
                box(title='Salary and Vacancies', width=12,status='warning',solidHeader = TRUE, 
                    highchartOutput("SalaryVacancyTree", height = "600px")%>%
                        helper(type = "inline",
                               title = "Salary and Vacancies",
                               content = "The Size of the box represents the number of vacancies of the company while the color intensity represents the average salary",
                               size = "s"))),
                
        ),

                
        ## Salary by Company
        tabItem(tabName = "Salary_by_Company",
                h2("Compare Salaries across Companies"),
                tags$h4("Compare the salary distribution of the different companies."),
                fluidRow(
                    box(width = 12,
                        title = "Salary",
                        color = "green", ribbon = TRUE, title_side = "top right",
                        column(width = 5),
                        plotlyOutput("salaryfig"),
                        selectizeInput('SalaryCompany','Select Companies:',c(sort(unique(salary_data$Company))),multiple = T, options = list(maxItems = 5, placeholder = "Select up to 5 companies"))
                    ))),
        
        #Ratings Content
        
        ## Ratings by Industry
        tabItem(tabName = "Ratings_by_Industry",
                h2("Compare Ratings across Industries"),
                tags$h4("Compare the ratings of the different industries."),
                box(title ='Type of Comparison', width=6,status='warning',
                    radioButtons('radar_ind_type','Type:',c("Industry_Average","Personal_Score"))%>%
                        helper(type = "inline",
                               title = "Select Type of Comparison",
                               content = "Choose to compare the ratings of the industries against the user score or the industry average",
                               size = "s")
                ),
                box(title = "Industries of Interest",width=6, status = 'warning',
                    selectizeInput('radarind','Select Industries:', Industry_list ,multiple = T, options = list(maxItems = 4, placeholder = "Select up to 4 Industries"))),
                box(title='Ratings', width=12,status='warning',solidHeader = TRUE, 
                    plotOutput("Radar_Ind", height = "550px")%>%
                        helper(type = "inline",
                               title = "Ratings",
                               content = "The red line represents the scores of the Industries while the shaded area represents the comparion type (either user score or industry average)",
                               size = "s")),
                fluidRow(
                    box(title = "Happiness ratings",width = 12,
                        plotlyOutput("happiness_fig"),
                        selectizeInput('HappinessIndustry','Select Industries:',c(sort(unique(happiness_count2$Industry))),multiple = T, options = list(maxItems = 5, placeholder = "Select up to 5 industries"))
                    )),
                    
                    
                ),
        ## Ratings by Company
        
        tabItem(tabName = "Ratings_by_Company",
                h2("Compare Ratings across Companies"),
                tags$h4("Compare the ratings of the different companies."),
                fluidRow(
                box(title ='Type of Comparison',width=6, status='warning', 
                    radioButtons('radar_comp_type','Type:',c("Companies_Average","Personal_Score"))%>%
                        helper(type = "inline",
                               title = "Select Type of Comparison",
                               content = "Choose to compare the ratings of the Companies against the user score or the Company average",
                               size = "s")
                ),
                box(title = "Companies of Interest", width=6,status = 'warning',
                    selectizeInput('radarcomp','Select Companies:', Company_list ,multiple = T, options = list(maxItems = 4, placeholder = "Select up to 4 Companies"))),
                box(title='Ratings', width=12, status='warning',solidHeader = TRUE, 
                    plotOutput("Radar_Comp", height = "450px")%>%
                        helper(type = "inline",
                               title = "Ratings",
                               content = "The red line represents the scores of the Companies while the shaded area represents the comparion type (either user score or industry average)",
                               size = "s")),
                    
                )),
        
        #Skills Content
        tabItem(tabName = "Skills",
                h2("Check Skills Required"),
                tags$h4("Check what are the technical and people skills required at the different companies."),
                fluidRow(
                    
                    box(title='Check Degree Requirements by Industry', status='warning',solidHeader = TRUE, plotlyOutput('degree_ind'),
                        radioButtons('industry_type','Select Type',c('Degrees','Subjects')),
                        selectInput('industry_degree','Select Industry',unique(industry$Industry))),
                    
                    
                    box(title='Check Degree Requirements by Company', status='danger',solidHeader = TRUE, plotlyOutput('degree_comp'),
                        radioButtons('company_type','Select Type',c('Degrees','Subjects')),
                        selectInput('company_degree','Select Company',unique(industry$Company))),
                    
                    
                    box(title='Check Industry Level Skills', status='primary',solidHeader = TRUE, plotlyOutput('Skills_ind'),
                        selectInput('industry_skill','Select Industry',unique(industry$Industry)),
                        sliderInput('industry_skill_freq','Select Number of Words',1,10,5)),
                    
                    box(title='Check Company Level Skills', status='danger',solidHeader = TRUE, plotlyOutput('Skills_comp'),
                        selectInput('company_skill','Select Company',unique(industry$Company)),
                        sliderInput('company_skill_freq','Select Number of Words',1,10,5))
                    
                )),
        
       
        #Reviews Content
        tabItem(tabName = "Reviews",
                h2("Check Company Reviews"),
                tags$h4("Sentiment Analysis of the reviews of Google"),
                
                fluidRow(
                    box(title="Most common positive and negative by Reviewer's Location",plotOutput('location_sentiment'),
                        selectInput('reviews_location','Company:',unique(review$Review_Location))),
                    
                    box(title="Most common positive and negative by Reviewer's Position",plotOutput('position_sentiment'),
                        selectInput('reviews_position','Company:',unique(review$Review_Author))),
                    
                    box(title='See most frequent words',width=6, status='primary',solidHeader = TRUE, plotOutput('wordcloud'),
                        selectInput('reviews','Company:','Google'),
                        radioButtons('rating','Rating:',c('1.0','2.0','3.0','4.0','5.0'),inline = TRUE),
                        sliderInput('word_freq','Number of Words:',5,100,60)),
                    
                    box(title='Company Emotional Sentiment Analysis',width=6, status='primary',solidHeader = TRUE, plotlyOutput('emotion_sentiment')))
        ),
        
        # More info content
        tabItem(
            tabName = "More_info", 
            
            navbarPage("More information",
                       
                       
                       tabPanel("About the application",
                                mainPanel(
                                    
                                    div(class="panel panel-default",
                                        div(class="panel-body",  
                                            fluidRow(
                                                shiny::HTML("<br><br><center> 
                                            <h1>More Information</h1> 
                                            </center>
                                            <br>
                                            <br>"),
                                                style = "height:100px;"),
                                            
                                            fluidRow(
                                                
                                                column(12,
                                                       tags$ul(align = "Center",
                                                               tags$span(h4("JobVis is designed to allow job seekers to build interactive visualizations and make useful comparisons between companies and industries based on job characteristics they value the most. Through these features, our app aims to provide deeper insights on an industry and company level so that users can assess whether they are aligned with the company or industry’s requirements and if they fit in with the company/industry culture.")), 
                                                               tags$span(h4("Motivation for this project stemmed from our observations that career portals very rarely provided a holistic view of culture, ratings, skills required (etc), at either an industry level or company level. Instead users would have to manually check each job posting to form an understanding of each characteristic. Additionally, we found that many lacked the nuances of allowing users to make tailored comparisons, and if they did it would be in a tabular format (example: Glassdoors Salary). Another major flaw was that portals did not display visualizations that allow users to keep track of their job applications and success rates. As such, JobVis addresses these pain points through graphical representations, filter functions and an improved UI layout that simplifies and adds value to the job searching experience.")), 
                                                               tags$span(h4("To note, we believe this app is particularly important today since Covid-19 has severely impacted the job market around the world, and has left job seekers questioning which industries and companies are still hiring, what skills they should upgrade to improve chances of landing jobs, and what industry/company best suits their requirements. Moreover, we see that companies are switching to utilizing AI to select candidates, thus it has become imperative for job seekers to determine the keywords that companies are searching for in each job posting and address them in either their resumes or in their job interviews."))
                                                       )
                                                )))))
                                
                       ),
                       tabPanel("About the Creators",
                                mainPanel(
                                    
                                    div(class="panel panel-default",
                                        div(class="panel-body",  
                                            fluidRow(
                                                shiny::HTML("<br><br><center> 
                                            <h1>About the Creators</h1> 
                                            </center>
                                            <br>
                                            <br>"),
                                                style = "height:100px;"),
                                            
                                            fluidRow(
                                                
                                                column(12,
                                                       tags$ul(align = "Center",
                                                               tags$span(h4("This app is created by Group 10 of DBA3702 Class SA1. The members include Chloe, Jovita, Saloni and Ying Jie."))
                                                       )))))
                                )),
                       tabPanel("What makes us unique",
                                mainPanel(
                                    fluidRow(
                                        column(6),
                                        column(12,
                                               # Panel for Background on Data
                                               div(class="panel panel-default",
                                                   div(class="panel-body",  
                                                       tags$div( align = "center",
                                                                 div( align = "center", 
                                                                      h4("What makes us unique?")
                                                                 )
                                                       ),
                                                       tags$p(h6("JobVis has three main categories of features that sets us apart from conventional job portals such as Symplicity, Indeed and Glassdoors. These features directly target pain points that we found job seekers find most frustrating. Thus, our unique selling points are as listed below:")),
                                                   ))
                                        )),
                                    
                                    
                                    fluidRow(
                                        box(title = "1. Fitting into the culture", width = 12, collapsible = T,collapsed = TRUE,
                                            box(tags$span(h6("Understanding the company/industry culture in terms of its overall review, happiness review, work-life balance, pay benefits, job security and advancement, etc, is pivotal in determining the user’s fit in the company/industry. Most job portals place very little importance on this or simply provide the statistics without any option to compare.  ", align = "left"))), 
                                            
                                            fluidRow(
                                                box(title = "Radar charts with comparing capabilities",collapsible = T, collapsed = T, tags$ul(
                                                    tags$span(h6("This feature creates radar charts that allow users to see how an industry perform against other industries in terms of several rating metrics. There are seven ratings altogether and they are (1) Overall Review, (2) Work Life Balance, (3) Pay Benefits, (4) Job Security and Advancement/Security , (5) Management, (6) Culture and (7) Woman Ratings.", align = "left")), 
                                                    
                                                )),
                                                box(title = "Reviews Analysis",collapsible = T, collapsed = T, tags$ul(
                                                    tags$span(h6("
Develop a comprehensive understanding of what employees are saying about working at the company! The word cloud represents the top positive and negative words users have mentioned, while the sentiment analysis graphs show the frequency of the different emotional words that employees have shown in their reviews", align = "left")), 
                                                    
                                                )))
                                        )),
                                    fluidRow(
                                        box(title = "2. Discovering skills requirement", width = 12, collapsible = T,collapsed = TRUE,
                                            
                                            fluidRow(
                                                box(title = "Degree requirements",collapsible = T, collapsed = T, tags$ul(
                                                    tags$span(h6("[Currently only available for Google] This feature allows users to determine the level of education the company is looking for for each of the role categories", align = "left")), 
                                                    
                                                )),
                                                box(title = "People skills requirements",collapsible = T, collapsed = T, tags$ul(
                                                    tags$span(h6("Determine what People skills (such as Decision Making, Communication, etc) companies are frequently mentioning in their job listings. For Google, the feature will allow users to further determine the skills required for each role category", align = "left")), 
                                                    
                                                )),
                                                box(title = "Technical skills requirements",collapsible = T, collapsed = T, tags$ul(
                                                    tags$span(h6("Determine what Technical skills (such as Machine Learning, Analytics, etc) companies are frequently mentioning in their job listings. For Google, the feature will allow users to further determine the skills required for each role category", align = "left")), 
                                                    
                                                )))
                                            
                                        )),
                                    fluidRow(
                                        box(title = "3. Discover trends in Demand and Compare Salaries", width = 12, collapsible = T,collapsed = TRUE,
                                            box(title = "Heat Maps to check the hottest hiring spots",collapsible = T, collapsed = T, tags$ul(
                                                tags$span(h6("This feature provides a visualization that shows the number of job vacancies across the different US states across different industries", align = "left")), 
                                                
                                            )),
                                            box(title = "Compare Salaries ",collapsible = T, collapsed = T, tags$ul(
                                                tags$span(h6("The dumbbell chart allows users to select five preferred companies to compare maximum and minimum salaries of each job role within the chosen companies and get a gauge of the range of prospective jobs they may be interested in. ", align = "left")), 
                                                
                                            ))
                                            
                                        )),
                                    fluidRow(
                                        box(title = "4. Track your Application Progress", width = 12, collapsible = T,collapsed = TRUE,
                                            tags$ul(
                                                tags$span(h4("Sankey Graph", align = "left")), 
                                                tags$span(h6("The Sankey Network helps the user track his jobs application progress. The nodes represent the actions of both the user and the company while the size of the link represents the number of actions linking each node (For example: Jobs applied to -> No replies 50 means that 50 of the companies the user applied to did not give a reply to the user)"))
                                                
                                            )
                                        )),
                                    fluidRow(
                                        box(title = "5. Other Considerations", width = 12, collapsible = T,collapsed = TRUE,
                                            tags$ul(
                                                tags$span(h4("CV Scanner", align = "left")), 
                                                tags$span(h6("The application will extract information on the skill sets and experiences listed in the uploaded CV and make comparisons against those required by the companies. We can use shinysense feature (that includes using camera) to add this feature into our app."))
                                                
                                            )
                                        ))
                                    
                                ),
                               )
            )
        )
        )))
#--------------------------------------------------------------------------------------------------------------------------------   
                
server <- function(input, output, session) {
    observe_helpers(withMathJax = TRUE)
    # Popup
    
    Intro_modal <- modalDialog(
        title = "Welcome to JobVis!", 
        HTML("JobVis is a job application that will allow you to make a well rounded and educated decision on which jobs you
        want to pursue by curating and visualizing important information about the various job listings on the market. <br> <br> 
        "),
        easyClose = T,
        size =  "m",
        footer = list(modalButton("Close")),
        
    )
    #Modal_2 <- modalDialog(
    #    title = "Title 2",
    #    "Message 2",
    #    easyClose = T,
    #    footer = list(modalButton("Close")),
    #)
    
    # Show the modal on start up
    showModal(Intro_modal)
    
    #Dashboard Content 
    output$valuebox_vacancies <- renderValueBox({
        valueBox(
            formatC(nrow(industry), format="d", big.mark=',')
            ,"job vacancies available"
            ,icon = icon("search")
            ,color = "red")  
    })
    
    output$valuebox_companies <- renderValueBox({
        valueBox(
            formatC(length(Company_list), format="d", big.mark=',')
            ,"companies to apply for"
            ,icon = icon("briefcase")
            ,color = "purple")  
    })
    
    output$valuebox_countries <- renderValueBox({
        valueBox(
            formatC(nrow(job_vacancies), format="d", big.mark=',')
            ,"states to choose from"
            ,icon = icon("globe")
            ,color = "yellow")  
    })
    
    output$valuebox_industries <- renderValueBox({
        valueBox(
            formatC(length(industries), format="d", big.mark=',')
            ,"industries to work at"
            ,icon = icon("stats",lib='glyphicon')
            ,color = "aqua")  
    })
    
    
    card_swipe <- callModule(shinyswipr, "my_swiper")
    sample_data <- industry[sample(nrow(industry), 1), ]
    output$job_title <- renderText({ sample_data$Title })
    output$company_name <- renderText({ sample_data$Company })
    output$job_industry <- renderText({ sample_data$Industry })
    output$job_location <- renderText({ sample_data$Location})
    
    # Render the table of swipes
    output$resultsTable <- renderDataTable({ appVals$swipes })
    
    # Set up reactive values object
    appVals <- reactiveValues(
        swipes = data.frame(
            job_title = character(), company_name = character(), job_industry=character(), job_location=character(),swipe = character()
        )
    )
    
    observeEvent(card_swipe(), {
        
        # Record the last swipe result
        latest_result <- data.frame(
            job_title = appVals$sample_data$Title,
            company_name = appVals$sample_data$Company,
            job_industry = appVals$sample_data$Industry,
            job_location = appVals$sample_data$Location,
            swipe = card_swipe())
        
        # Add to table of all swipe results
        appVals$swipes <- rbind(latest_result, appVals$swipes)
        
        # Send results to the output
        output$resultsTable <- renderTable({ appVals$swipes })
        
        # Update the villager
        appVals$sample_data <- sample_n(industry, 1)
        
        # Send update to ui
        output$job_title <- renderText({ appVals$sample_data$Title })
        output$company_name <- renderText({ appVals$sample_data$Company })
        output$job_industry <- renderText({ appVals$sample_data$Industry})
        output$job_location <- renderText({ appVals$sample_data$Location})
        
        the_data <- eventReactive(
            { card_swipe() }
        )
        
        output$table <- renderTable(the_data())
        
    }) 
    
    #For the filtered table
    subset_summary <- industry%>%select(Title,State_Name,Company,Industry,Salary_Min,Salary_Max,Overall_Review)
    
    filtered_location <- reactive({
        if(input$filter_location=="All Locations"){
            subset_summary
        }
        else{subset_summary()%>%filter(State_Name == input$filter_location)}
    })
    
    filtered_company <- reactive({
        if(input$filter_company=="All Companies"){
            filtered_location()
        }
        else{filtered_location()%>%filter(Company == input$filter_company)}
    })
    
    filtered_industry <- reactive({
        if(input$filter_industry=="All Industries"){
            filtered_company()
        }
        else{filtered_company()%>%filter(Industry == input$filter_industry)}
    })
    
    filtered_salary_review <- reactive({
        filtered_industry()%>%filter(Overall_Review>=input$filter_rating & Salary_Min >= input$filter_salary[1])
    })
    
    
    output$job_summary_table <- renderTable({
        filtered_salary_review()
    })
    
    #Info Boxes at the top of dashboard
    output$out_salary <- renderInfoBox({
        infoBox("Click to open tab",  
                a("Salary Comparison Tool", onclick = "openTab('Salary_by_Company')", href="#"),
                icon = icon('usd', lib = 'glyphicon'), color = "green"
        )
    })
    
    output$out_requirements <- renderInfoBox({
        infoBox("Click to open tab",  
                a("Skills Requirement Checker", onclick = "openTab('Skills')", href="#"),
                icon = icon('check', lib = 'glyphicon'), color = "green"
        )
    })
    
    output$out_jobfilter <- renderInfoBox({
        infoBox("Click to open tab",  
                a("Job Filter Toolkit", onclick = "openTab('Jobs_by_Company')", href="#"),
                icon = icon('filter', lib = 'glyphicon'), color = "green"
        )
    })
    
    
    date_posted <- industry %>% group_by(Date)%>%summarise(Days_Ago=n())
    date_posted$Date <- as.numeric(date_posted$Date)
    date_posted <- date_posted[complete.cases(date_posted$Date),]
    date_posted <- date_posted %>% arrange(desc(Date))
    output$job_posting <-renderPlotly({plot_ly(
        x = date_posted$Date,
        y = date_posted$Days_Ago, text=date_posted$Days_Ago,textposition='auto',
        type = "scatter", mode='lines+markers', marker = list(color = 'rgb(158,202,225)',
                                                              line = list(color = 'rgb(8,48,107)')))%>% layout(title= "Recent Job Posting Activity Rate",
                                                                                                               yaxis=list(title="No. of Job Postings"),
                                                                                                               xaxis=list(title="Date Posted (Days Ago)",autorange="reversed"))
    })
        
    
    #User's Profile Content
    # Sankey graph
    
    links <- data.frame(
        source=c("Jobs_applied_to","Jobs_applied_to", "Job_applied_to", "Replies", "Replies", "Initial Interview",
                 "Initial Interview","Initial Interview", "Initial Interview", "Offer Received", "Offer Received", "Final Interview",
                 "Final Interview","Final Interview", "Offer Received ", "Offer Received "), 
        target=c("Rejected by Company","No replies", "Replies", "Rejected by Company ", "Initial Interview", "Rejected by Company  ",
                 "No Followup", "Offer Received", "Final Interview", "Applicant Accepts", "Applicant Rejects", "Rejected by Company   ",
                 "No Followup " , "Offer received ", "Applicant Accepts ", "Applicant Rejects "),
        value=c(25 ,50 ,25 ,6 ,19 ,10 ,2 ,2 ,5 ,0 ,2, 2, 1, 2, 1, 1 )
    )
    
    
    # From these flows we need to create a node data frame: it lists every entities involved in the flow
    nodes <- data.frame(
        name=c(as.character(links$source), 
               as.character(links$target)) %>% unique()
    )
    
    # With networkD3, connection must be provided using id, not using real name like in the links dataframe.. So we need to reformat it.
    links$IDsource <- match(links$source, nodes$name)-1 
    links$IDtarget <- match(links$target, nodes$name)-1
    
    # Make the Network
    sankey_plot <- sankeyNetwork(Links = links, Nodes = nodes,
                                 Source = "IDsource", Target = "IDtarget",
                                 Value = "value", NodeID = "name", 
                                 sinksRight=FALSE, fontSize = 12)
    output$UserProgress <- renderSankeyNetwork({sankey_plot})
   
    #Jobs Content
    #Jobs by Industry Content
    ## Create a summary grouped by State/State+Industry
    vac_state_data <- industry %>% group_by(State_Name) %>% summarize(number_of_vacancies = n())
    colnames(vac_state_data) <- c("state","number_of_vacancies" )
    vac_ind_state_data <- industry %>% group_by(State_Name, Industry) %>% summarize(number_of_vacancies = n())
    colnames(vac_ind_state_data) <- c("state", "Industry","number_of_vacancies" )
    output$heatmapUI <- renderUI({
       plotOutput('Job_Industry_State',height=620)
        }
    )
    output$Job_Industry_State <- renderPlot({
        if(input$usmap_ind =="All"){
            plot_usmap(data = vac_state_data, values = "number_of_vacancies", color = "red", labels= TRUE) + 
                scale_fill_continuous(low = "white", high = "red", name = "Number of Vacancies", label = scales::comma
                ) + theme(legend.position = "right")}
        else{
            vac_ind_state_data1 <- vac_ind_state_data  %>% filter(Industry==input$usmap_ind)
            plot_usmap(data = vac_ind_state_data1,values = "number_of_vacancies", color = "red", labels= TRUE) + 
                scale_fill_continuous(low = "white", high = "red", name = "Number of Vacancies", label = scales::comma
                ) + theme(legend.position = "right")
        }
        
        
    })
    # Jobs By Company
    thedata <- reactive({
        if (input$country=="All Countries"){
            job_vacancies_complete%>% 
                mutate(iso3 = countrycode(Country,"country.name","iso3c")) %>%
                filter(VacanciesCount>=input$vacancies)}
        else{
            job_vacancies_complete %>%
                filter(Country==input$country) %>%
                mutate(iso3 = countrycode(Country,"country.name","iso3c")) %>%
                filter(VacanciesCount>=input$vacancies)}
    })
    
    #Plotting the choropleth map 
    output$plotChoroplethMap <-renderHighchart(highchart(type = "map") %>% 
                                               hc_add_series_map(map = worldgeojson, df = thedata(), value = "Num_Vacancy",joinBy ="iso3" ) %>% 
                                               hc_colorAxis(stops = color_stops()) %>% 
                                               hc_tooltip(useHTML=TRUE,headerFormat='',pointFormat = paste0('  {point.Country} Vacancies : {point.VacanciesCount} ')) %>% 
                                               hc_title(text = 'Map of Job Vacancies around the World') %>% 
                                               hc_subtitle(text = paste0('Vacancies more than: ',input$vacancies)) %>% hc_exporting(enabled = TRUE,filename = 'custom'))
    
                                                   
    #Plotting the sunburst chart    
    getSunburst <- reactive({
        continent_countries <- gapminder[gapminder$continent == input$continent,]
        df3_continent <- google[google$Country %in% continent_countries$country,]
        df3_sunburst <- data.frame(labels=character(),values=numeric(),parents=character(),ids=character())
        df3_sunburst[1,]<- c(input$continent,nrow(df3_continent),"",input$continent)
        if (input$filter_type=="Country"){
            df3_x <- df3_continent%>%group_by(Country)%>%summarise(Count=n())
            df3_x$parents <- input$continent
            df3_x$ids <- paste0(df3_x$parents," - ",df3_x$Country)
            df3_x <- df3_x %>%rename(labels=Country,values=Count)
            df3_y <- df3_continent%>%group_by(Category,Country)%>%summarise(values=n())
            df3_y$parents <- paste0(input$continent," - ",df3_y$Country)
            df3_y$ids <- paste0(df3_y$parents," - ",df3_y$Category)
            df3_y  <- df3_y %>%rename(labels=Category)
            df3_y <-df3_y %>%select(-Country)}
        else if(input$filter_type=="Category"){
            df3_x <- df3_continent%>%group_by(Category)%>%summarise(Count=n())
            df3_x$parents <- input$continent
            df3_x$ids <- paste0(df3_x$parents," - ",df3_x$Category)
            df3_x <- df3_x %>%rename(labels=Category,values=Count)
            df3_y <- df3_continent%>%group_by(Country,Category)%>%summarise(values=n())
            df3_y$parents <- paste0(input$continent," - ",df3_y$Category)
            df3_y$ids <- paste0(df3_y$parents," - ",df3_y$Country)
            df3_y <- df3_y %>%rename(labels=Country)
            df3_y <-df3_y %>%select(-Category)}
        
        df3_sunburst<- rbind(df3_sunburst,df3_x,df3_y)
        df3_sunburst
        
    })
    
    output$plotSunburst <- renderPlotly({
        plot_ly(
            data = getSunburst(), ids = ~ids, 
            labels= ~labels, parents = ~parents, 
            values= ~values, type='sunburst', branchvalues = 'total') %>%
            layout(plot_bgcolor  = "rgba(0, 0, 0, 0)",
                   paper_bgcolor = "rgba(0, 0, 0, 0)",
                   fig_bgcolor   = "rgba(0, 0, 0, 0)")
    })
    #Salary Content 
    
    ## Salary by Industry 
    # Treemap for salary/vacancy
    
    ## Create salary/vacancy summary by company/ company + industry
    vac_sal_comp <- industry %>% group_by(Company) %>% summarize(number_of_vacancies = n(), mean_salary = mean(Salary_Avg))
    colnames(vac_sal_comp) <- c("Company","number_of_vacancies","Average_Salary" )
    vac_sal_comp_ranked <- head(arrange(vac_sal_comp, desc(number_of_vacancies)), n=50)
    vac_sal_comp_ind <- industry %>% group_by(Company, Industry) %>% summarize(number_of_vacancies = n(), mean_salary = mean(Salary_Avg))
    colnames(vac_sal_comp_ind) <- c("Company","Industry","number_of_vacancies" ,"Average_Salary" )
    ## Create a summary grouped by State/State+Industry
    vac_state_data <- industry %>% group_by(State_Name) %>% summarize(number_of_vacancies = n())
    colnames(vac_state_data) <- c("state","number_of_vacancies" )
    vac_ind_state_data <- industry %>% group_by(State_Name, Industry) %>% summarize(number_of_vacancies = n())
    colnames(vac_ind_state_data) <- c("state", "Industry","number_of_vacancies" )
    output$SalaryVacancyTree <- renderHighchart({
        if(input$salary_tree_ind =="All"){
            hctreemap2(vac_sal_comp_ranked, group_vars = c("Company"),size_var = "number_of_vacancies", color_var = "Average_Salary",
                       layoutAlgorithm = "squarified")}
        else{hctreemap2(vac_sal_comp_ind %>% filter(Industry == input$salary_tree_ind), 
                        group_vars = c("Company"),
                        size_var = "number_of_vacancies", color_var = "Average_Salary",
                        layoutAlgorithm = "squarified")}
        
    })
    
    ## Salary by Company
    output$salaryfig <- renderPlotly({plot_ly(salary_data[salary_data$Company %in% input$SalaryCompany ,],color = I("gray80")) %>% 
            add_segments(x = ~Salary_Min, xend = ~Salary_Max, y = ~Company, yend = ~Company, showlegend = FALSE) %>% 
            add_markers(x = ~Salary_Min, y = ~Company, name = "Min", color = I("pink"),text = ~Title, hovertemplate = '<b>Title</b>: %{text}<br><b>Salary</b>: %{x}') %>% 
            add_markers(x = ~Salary_Max, y = ~Company,  name = "Max", color = I("blue"),text = ~Title, hovertemplate = '<b>Title</b>: %{text}<br><b>Salary</b>: %{x}') %>% 
            layout(title = "Earnings range per company", xaxis = list(title = "Monthly Salary (in thousands)"), yaxis = list(title = paste0(c(rep("&nbsp;",25),"Company", rep("&nbsp;", 50), rep("\n&nbsp;", 10)), collapse = ""))
            ) 
    })
    
    
    #Ratings Content
    # Ratings by Industry Content
    
    ## Create data for radar plot for ratings by industry
    df_radar_ind <- industry %>% filter(Industry != "") %>% group_by(Industry) %>% summarize(Overall = mean(Overall_Review, na.rm = T), 
                                                                                             Balance = mean(Work_Life_Balance, na.rm = T), Pay = mean(Pay_Benefits, na.rm = T), Stability = mean(Job_Security_and_Advancement,na.rm = T),
                                                                                             Management = mean(Management, na.rm = T), Culture = mean(Culture, na.rm = T), Woman = mean(Woman_Rating, na.rm = T))
    df_radar_ind1 <-df_radar_ind[,-1]
    rownames(df_radar_ind1)<-df_radar_ind$Industry
    
    ## Fetch minima and maxima of every column of the data-set
    colMax <- function (x) { apply(x, MARGIN=c(2), max, na.rm=T) }
    colMin <- function (x) { apply(x, MARGIN=c(2), min, na.rm=T) }
    maxmin_ind <- data.frame(max=colMax(df_radar_ind1),min=colMin(df_radar_ind1))
    ratings_max <- c(rep(5,7))
    ratings_min <- c(rep(0,7))
    maxmin_ratings <- as.data.frame(matrix(NA,nrow=7,ncol=2))
    maxmin_ratings[,1] <- ratings_min
    maxmin_ratings[,2] <- ratings_max
    rownames(maxmin_ratings) <-c("Overall" , "Balance" , "Pay", 'Stability', "Management", "Culture", "Woman")
    colnames(maxmin_ratings) <-c("min", "max")
    ## Calculate the average profile
    average_ind <- data.frame(rbind(maxmin_ind$max,maxmin_ind$min,t(colMeans(df_radar_ind1, na.rm=T))))
    colnames(average_ind) <- colnames(df_radar_ind1)
    
    output$Radar_Ind <- renderPlot({
        if (input$radar_ind_type == "Industry_Average") {
            data <- df_radar_ind1 %>% filter(row.names(df_radar_ind1) %in% input$radarind)
            par(mar=rep(1.6,4))
            par(mfrow=c(1,4))
            for (i in 1:nrow(data)) {
                toplot <- rbind(
                    maxmin_ratings$max,
                    maxmin_ratings$min,
                    average_ind[3,],
                    data[i,]
                )
                plot <- radarchart(
                    toplot,
                    axistype = 2,
                    pfcol = c(rgb(0.2,0.5,0.5,0.4),NA),
                    pcol= c(NA,2),
                    pty = 32, 
                    plty = 1,
                    plwd = 2,
                    title = row.names(data)[i]
                )
            }
            plot
            par(mfrow = c(1,1))
        }
        else {
            User_Results <- c(3.2, 3.0, 4.4, 3.6, 4.8, 3.4, 3)
            user_data <- data.frame(matrix(nrow=2,ncol=7))
            user_data[1,] <- User_Results
            colnames(user_data) <- c("Overall" , "Balance" , "Pay", 'Stability', "Management", "Culture", "Woman")
            
            
            
            data <- df_radar_ind1 %>% filter(row.names(df_radar_ind1) %in% input$radarind)
            par(mar=rep(1.6,4))
            par(mfrow=c(1,4))
            for (i in 1:nrow(data)) {
                toplot <- rbind(
                    maxmin_ratings$max,
                    maxmin_ratings$min,
                    user_data[1,],
                    data[i,]
                )
                plot <- radarchart(
                    toplot,
                    axistype = 2,
                    pfcol = c(rgb(0.8,0.2,0.5,0.4),NA),
                    pcol= c(NA,2),
                    pty = 32, 
                    plty = 1,
                    plwd = 2,
                    title = row.names(data)[i]
                )
            }
            plot
            par(mfrow = c(1,1))            
            
            
        }
        
        
    })
    
    output$happiness_fig <- renderPlotly({plot_ly(happiness_count2[happiness_count2$Industry %in% input$HappinessIndustry ,],color = I("gray80")) %>% 
            add_bars(x=~n, y = ~Industry, color = ~happiness_labels, orientation = 'h' ) %>% 
            layout(barmode='stack', xaxis=list( showgrid=T, showlegend = F, zeroline=T,showline=FALSE, ticks='',showticklabels=T))
    })
    
    
    #Ratings by Company Content
    
    # Radar Chart for Companies
    
    ## Create data for radar plot for ratings by Company
    df_radar_comp <- industry %>% filter(Company != "") %>% group_by(Company) %>% summarize(Overall = mean(Overall_Review, na.rm = T), 
                                                                                            Balance = mean(Work_Life_Balance, na.rm = T), Pay = mean(Pay_Benefits, na.rm = T), Stability = mean(Job_Security_and_Advancement,na.rm = T),
                                                                                            Management = mean(Management, na.rm = T), Culture = mean(Culture, na.rm = T), Woman = mean(Woman_Rating, na.rm = T))
    df_radar_comp1 <-df_radar_comp[,-1]
    rownames(df_radar_comp1)<-df_radar_comp$Company
    # Fetch minima and maxima of every column of the data-set
    maxmin_comp <- data.frame(max=colMax(df_radar_comp1),min=colMin(df_radar_comp1))
    
    # Calculate the average profile
    average_comp <- data.frame(rbind(maxmin_comp$max,maxmin_comp$min,t(colMeans(df_radar_comp1, na.rm=T))))
    colnames(average_comp) <- colnames(df_radar_comp1)
    
    output$Radar_Comp <- renderPlot({
        if (input$radar_comp_type == "Companies_Average") {
            
            data <- df_radar_comp1 %>% filter(row.names(df_radar_comp1) %in% input$radarcomp)
            par(mar=rep(1.6,4))
            par(mfrow=c(1,4))
            for (i in 1:nrow(data)) {
                toplot <- rbind(
                    maxmin_ratings$max,
                    maxmin_ratings$min,
                    average_comp[3,],
                    data[i,]
                )
                plot <- radarchart(
                    toplot,
                    axistype = 2,
                    pfcol = c(rgb(0.2,0.5,0.5,0.4),NA),
                    pcol= c(NA,2),
                    pty = 32, 
                    plty = 1,
                    plwd = 2,
                    title = row.names(data)[i]
                )
            }
            plot
            par(mfrow = c(1,1))
        }
        else {
            User_Results <- c(3.2, 3.0, 4.4, 3.6, 4.8, 3.4, 3)
            user_data <- data.frame(matrix(nrow=2,ncol=7))
            user_data[1,] <- User_Results
            colnames(user_data) <- c("Overall" , "Balance" , "Pay", 'Stability', "Management", "Culture", "Woman")
            
            
            data <- df_radar_comp1 %>% filter(row.names(df_radar_comp1) %in% input$radarcomp)
            par(mar=rep(1.6,4))
            par(mfrow=c(1,4))
            for (i in 1:nrow(data)) {
                toplot <- rbind(
                    maxmin_ratings$max,
                    maxmin_ratings$min,
                    user_data[1,],
                    data[i,]
                )
                plot <- radarchart(
                    toplot,
                    axistype = 2,
                    pfcol = c(rgb(0.8,0.2,0.5,0.4),NA),
                    pcol= c(NA,2),
                    pty = 32, 
                    plty = 1,
                    plwd = 2,
                    title = row.names(data)[i]
                )
            }
            plot
            par(mfrow = c(1,1))            
            
            
        }
        
        
    })
    
    
    #Skills Content
    softskills <- c('integrity','dependability','effective communication','communication','teamwork','creativity','problem solving','critical','adaptability','organization','willingness to learn','empathy','humble','leadership','lead','resourceful','analytical','motivated','interpersonal','curious','driven','research','work ethic','flexibility','attention-to-detail','confidence','presentation','training','management','independent','dedicated','social','enthusiastic','collaboration','passionate','communicate','decision making','quick','friendly','networking','relationship','team building','analytical thinker','storytelling','critical observation','will learn','innovation','logical thinking','troubleshooting','conflict resolution','deal making','inspiring people','mentoring','resolving issues','managing difficult conversations','giving clear feedback','dealing with politics','hard working','business ethics','meeting deadlines','multitasking','persistence','proper business etiquette','planning','reliability','self directed','experience')
    hardskills <- c("analytic solutions","machine learning","predictive modeling","database systems","clinical decision engines", "algorithms", "NLP/ML", "SQL",  "MongoDB","DynamoDB", "R, ","Python","dplyr","GGPlot", "Pandas","OLS","MLE","Machine Learning",  "Decision Tree","Random Forest","AI" , "Visualization","A/B tests set-up","Reporting",  "data visualizations","numpy", "scipy","scikit-learn", "tensorflow","pytorch" , "keras","genism", "vowpal wabbit","Heap.io","Google Analytics","Big Data","Business Analytics","Oracle","Relational Database Management System (RDMS)","Statistical Programming Language","Regression","Decision Trees","K-Means","Tableau","looker","R Programming" ,"Microsoft Office" , "SPSS","No-SQL", "Cassandra","Hadoop", "Pig","Hive", "HPCC Systems","Javascript" , "Java programming","PowerBI","Linux","TensorFlow", "Keras","Shiny","Artificial Intelligence","NLP", "Tesseract","Jenkins CI/CD", "Azure","logistic regression","k-means clustering","decision forests", "JavaScript","Cloud data", "MATLAB","Excel", "Jupyter","Gurobi","agile", "Git","Github" , "Qlikview","Business Intelligence", "supply chain","D3", "big data",'business sense','C',' API', 'Get Requests', 'Push Requests', 'Update Requests','AWS', 'Sagemaker','Power BI','PowerBI','Cognos','suite', 'Business Objects','Amplitude','Mixpanel','Salesforce', 'Qlik','Microstrategy', 'java','business analytics','reporting','C+','C++','perl','php','hadoop','web scraping','text mining','ruby','Word','NLP','unstructured data') %>% tolower()
    industry_unnest <- industry %>% select(Industry,Company, Description) %>% unnest_tokens(word,Description) %>% anti_join(.,stopwords,by='word') %>% filter(!str_detect(word,'[0-9]+'))
    google_unnest <- google %>% select(Category,`Minimum Qualifications`,`Preferred Qualifications`) %>% gather(.,Type,Qualifications,`Minimum Qualifications`:`Preferred Qualifications`) %>% select(-Type) %>% unnest_tokens(word,Qualifications) %>% anti_join(.,stopwords,by='word') %>% filter(!str_detect(word,'[0-9]+'))
    
    degrees <- c('ba','bs',"bachelor's",'bachelors','phd','masters','ms','mba',"master's","undergraduate","graduate")
    subjects <- c('engineering','marketing','finance','Arts','Accounting','Law','economics','science','humanities') %>% tolower()
    
    output$degree_ind <- renderPlotly({
        if(input$industry_type=='Degrees') {
            industry_unnest %>% filter(Industry==input$industry_degree) %>% filter(word %in% degrees) %>% 
                mutate(word=case_when(word=='bs'~'Bachelors',word=='ba'~'Bachelors',word=="bachelor's"~'Bachelors',word=="bachelors"~"Bachelors", word=='mba'~'MBA',word=='ms'~'Masters',word=="master's"~'Masters',word=="masters"~'Masters',word=='phd'~'PhD',word=='undergraduate'~'Bachelors',word=='graduate'~'Graduates',TRUE~word)) %>% count(word,sort=TRUE) %>% 
                mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% plot_ly(x=~case_when(word=='PhD'~'PhD', word=='MBA'~'MBA',TRUE~str_to_title(word)),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Degrees'), yaxis=list(visible=FALSE))
        } else {
            industry_unnest %>% filter(Industry==input$industry_degree) %>% filter(word %in% subjects) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>%
                plot_ly(x=~str_to_title(word),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Subject'), yaxis=list(visible=FALSE))}
    })
    
    output$degree_comp <- renderPlotly({
        if(input$company_type=="Degrees"){
            if(input$company_degree=='Google') {
                google_unnest %>% filter(word %in% degrees) %>% mutate(word=case_when(word=='bs'~'Bachelors',word=='ba'~'Bachelors',word=="bachelor's"~'Bachelors',word=="bachelors"~"Bachelors", word=='mba'~'MBA',word=='ms'~'Masters',word=="master's"~'Masters',word=="masters"~'Masters',word=='phd'~'PhD',word=='undergraduate'~'Bachelors',word=='graduate'~'Graduates',TRUE~word)) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% 
                    plot_ly(x=~case_when(word=='PhD'~'PhD',word=='MBA'~'MBA',TRUE~str_to_title(word)),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Degrees'), yaxis=list(visible=FALSE))
            }else{
                industry_unnest %>% filter(Company==input$company_degree) %>% filter(word %in% degrees) %>% mutate(word=case_when(word=='bs'~'Bachelors',word=='ba'~'Bachelors',word=="bachelor's"~'Bachelors',word=="bachelors"~"Bachelors", word=='mba'~'MBA',word=='ms'~'Masters',word=="master's"~'Masters',word=="masters"~'Masters',word=='phd'~'PhD',word=='undergraduate'~'Bachelors',word=='graduate'~'Graduates',TRUE~word)) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% 
                    plot_ly(x=~case_when(word=='PhD'~'PhD',word=='MBA'~'MBA',TRUE~str_to_title(word)),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Degrees'), yaxis=list(visible=FALSE))}
        } else {
            if(input$company_degree=='Google') {
                google_unnest%>% filter(word %in% subjects) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% plot_ly(x=~str_to_title(word),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Subject'), yaxis=list(visible=FALSE))
            } else {
                industry_unnest %>% filter(Company==input$company_degree) %>% filter(word %in% subjects) %>% count(word,sort=TRUE) %>% mutate(percentage=paste(formatC((n/sum(n)*100),2),'%')) %>% 
                    plot_ly(x=~str_to_title(word),y=~n,hovertext=~percentage) %>% layout(xaxis=list(title='Subject'), yaxis=list(visible=FALSE))
            }
        }
        
    })
    
    output$Skills_ind <- renderPlotly({
        
        industry_soft <-industry_unnest %>% filter(Industry==input$industry_skill) %>% filter(word %in% softskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% 
            top_n(input$industry_skill_freq,n) %>% plot_ly(x=~str_to_title(word),y=~n,type='bar', hovertext=~percentage, marker=list(color='purple'),showlegend=FALSE) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='People Skills'))
        
        industry_hard <-industry_unnest %>% filter(Industry==input$industry_skill) %>% filter(word %in% hardskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(input$industry_skill_freq,n) %>%
            plot_ly(type='bar',x=~case_when(word=='nlp'~'NLP',word=='sql'~'SQL',word=='aws'~'AWS',TRUE~str_to_title(word)),y=~n, hovertext=~percentage,marker = list(color="rgba(255, 0, 0, 0.6)")) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='Technical Skills'),showlegend=FALSE) 
        
        subplot(industry_soft,industry_hard,nrows=1,titleX = TRUE, titleY = TRUE)
        
    })
    
    output$Skills_comp <- renderPlotly({
        
        if(input$company_skill=='Google') {
            company_soft <- google_unnest %>% filter(word %in% softskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(input$company_skill_freq,n) %>% 
                plot_ly(type='bar',x=~str_to_title(word),y=~n, hovertext=~percentage,marker = list(color="purple"),showlegend=FALSE) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='People Skills'))
            
            company_hard <- google_unnest %>% filter(word %in% hardskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(input$company_skill_freq,n) %>% 
                plot_ly(type='bar',x=~case_when(word=='nlp'~'NLP',word=='sql'~'SQL',word=='aws'~'AWS',TRUE~str_to_title(word)),y=~n, hovertext=~percentage,marker = list(color="rgba(255, 0, 0, 0.6)"),showlegend=FALSE) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='Technical Skills'))
            
            subplot(company_soft,company_hard,nrows=1,titleX = TRUE, titleY = TRUE)
            
        } else {
            
            company_soft <-industry_unnest %>% filter(Company==input$company_skill) %>% filter(word %in% softskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(input$company_skill_freq,n) %>% 
                plot_ly(type='bar',x=~str_to_title(word),y=~n, hovertext=~percentage, marker=list(color='purple'),showlegend=FALSE) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='People Skills'))
            
            company_hard <-industry_unnest %>% filter(Company==input$company_skill) %>% filter(word %in% hardskills) %>% count(word,sort=TRUE) %>% mutate(percentage=paste0((formatC(((n/sum(n))*100),4)),'%')) %>% top_n(input$company_skill_freq,n) %>% 
                plot_ly(type='bar',x=~case_when(word=='nlp'~'NLP',word=='sql'~'SQL',word=='aws'~'AWS',TRUE~str_to_title(word)),y=~n, hovertext=~percentage,marker = list(color="rgba(255, 0, 0, 0.6)"),showlegend=FALSE) %>% layout(yaxis=list(visible=FALSE),xaxis=list(title='Technical Skills'))
            
            subplot(company_soft,company_hard,nrows=1,titleX = TRUE, titleY = TRUE)
        }})
    
    
    
    #Reviews Content 
    review_unnest <- review %>% select(-Company) %>% unnest_tokens(word,Review) %>% anti_join(.,stopwords, by='word') 
    
    output$wordcloud <- renderPlot({
        review_unnest %>% filter(Rating==input$rating) %>% inner_join(get_sentiments("bing")) %>% count(word, sentiment, sort = TRUE) %>%acast(word ~ sentiment, value.var = "n", fill = 0) %>% comparison.cloud(colors = c("red", "darkGreen"),max.words = input$word_freq, scale=c(4,0.5))
    })
    
    output$emotion_sentiment <- renderPlotly({
        review_unnest %>% inner_join(get_sentiments('nrc') )%>% count(sentiment)%>% mutate(percentage=paste(formatC((n/sum(n))*100,3),'%')) %>% 
            group_by(sentiment) %>% arrange(desc(n)) %>% top_n(10,n) %>% mutate(Type=ifelse(sentiment %in% c('fear','disgust','anger','negative','sadness'),'Negative Emotion','Positive Emotion')) %>% plot_ly(x=~str_to_title(sentiment),y=~n,hovertext=~percentage,color=~Type,colors=c('red','darkGreen')) %>% layout(xaxis=list(title='Sentiment'),yaxis=list(visible=FALSE))
    })
    
    output$location_sentiment <- renderPlot({
        review_unnest %>% inner_join(get_sentiments("bing")) %>% anti_join(stopwords) %>% mutate(word=lemmatize_words(word),word=case_when(word=='ambiguity'~'ambiguous',word=='effectively'~'effective',TRUE~word)) %>% filter(Review_Location==input$reviews_location) %>% count(word, sentiment, sort = TRUE) %>% group_by(sentiment) %>% 
            top_n(10,n) %>% ungroup() %>% mutate(word = reorder(word, n)) %>% ggplot(aes(str_to_title(word), n, fill = sentiment)) +geom_col(show.legend = FALSE) + facet_wrap(~sentiment, scales = "free_y") +labs(y = "Contribution to sentiment",x = NULL) +coord_flip()
    })
    
    output$position_sentiment <- renderPlot({
        review_unnest %>% inner_join(get_sentiments("bing")) %>% anti_join(stopwords) %>% mutate(word=lemmatize_words(word),word=case_when(word=='ambiguity'~'ambiguous',word=='effectively'~'effective',TRUE~word)) %>% filter(Review_Author==input$reviews_position) %>% count(word, sentiment, sort = TRUE) %>% group_by(sentiment) %>%
            top_n(10,n) %>% ungroup() %>% mutate(word = reorder(word, n)) %>% ggplot(aes(str_to_title(word), n, fill = sentiment)) +geom_col(show.legend = FALSE) + facet_wrap(~sentiment, scales = "free_y") +labs(y = "Contribution to sentiment",x = NULL) +coord_flip()
    })
    
    
}    

shinyApp(ui, server)